linux_old1/drivers/acpi/ec.c

1392 lines
37 KiB
C
Raw Normal View History

/*
* ec.c - ACPI Embedded Controller Driver (v3)
*
* Copyright (C) 2001-2015 Intel Corporation
* Author: 2014, 2015 Lv Zheng <lv.zheng@intel.com>
* 2006, 2007 Alexey Starikovskiy <alexey.y.starikovskiy@intel.com>
* 2006 Denis Sadykov <denis.m.sadykov@intel.com>
* 2004 Luming Yu <luming.yu@intel.com>
* 2001, 2002 Andy Grover <andrew.grover@intel.com>
* 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com>
* Copyright (C) 2008 Alexey Starikovskiy <astarikovskiy@suse.de>
*
* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or (at
* your option) any later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc.,
* 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
*
* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*/
/* Uncomment next line to get verbose printout */
/* #define DEBUG */
#define pr_fmt(fmt) "ACPI : EC: " fmt
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/types.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <linux/list.h>
#include <linux/spinlock.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
ACPI: Clean up inclusions of ACPI header files Replace direct inclusions of <acpi/acpi.h>, <acpi/acpi_bus.h> and <acpi/acpi_drivers.h>, which are incorrect, with <linux/acpi.h> inclusions and remove some inclusions of those files that aren't necessary. First of all, <acpi/acpi.h>, <acpi/acpi_bus.h> and <acpi/acpi_drivers.h> should not be included directly from any files that are built for CONFIG_ACPI unset, because that generally leads to build warnings about undefined symbols in !CONFIG_ACPI builds. For CONFIG_ACPI set, <linux/acpi.h> includes those files and for CONFIG_ACPI unset it provides stub ACPI symbols to be used in that case. Second, there are ordering dependencies between those files that always have to be met. Namely, it is required that <acpi/acpi_bus.h> be included prior to <acpi/acpi_drivers.h> so that the acpi_pci_root declarations the latter depends on are always there. And <acpi/acpi.h> which provides basic ACPICA type declarations should always be included prior to any other ACPI headers in CONFIG_ACPI builds. That also is taken care of including <linux/acpi.h> as appropriate. Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Matthew Garrett <mjg59@srcf.ucam.org> Cc: Tony Luck <tony.luck@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Acked-by: Bjorn Helgaas <bhelgaas@google.com> (drivers/pci stuff) Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> (Xen stuff) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-12-03 08:49:16 +08:00
#include <linux/acpi.h>
#include <linux/dmi.h>
ACPI: Clean up inclusions of ACPI header files Replace direct inclusions of <acpi/acpi.h>, <acpi/acpi_bus.h> and <acpi/acpi_drivers.h>, which are incorrect, with <linux/acpi.h> inclusions and remove some inclusions of those files that aren't necessary. First of all, <acpi/acpi.h>, <acpi/acpi_bus.h> and <acpi/acpi_drivers.h> should not be included directly from any files that are built for CONFIG_ACPI unset, because that generally leads to build warnings about undefined symbols in !CONFIG_ACPI builds. For CONFIG_ACPI set, <linux/acpi.h> includes those files and for CONFIG_ACPI unset it provides stub ACPI symbols to be used in that case. Second, there are ordering dependencies between those files that always have to be met. Namely, it is required that <acpi/acpi_bus.h> be included prior to <acpi/acpi_drivers.h> so that the acpi_pci_root declarations the latter depends on are always there. And <acpi/acpi.h> which provides basic ACPICA type declarations should always be included prior to any other ACPI headers in CONFIG_ACPI builds. That also is taken care of including <linux/acpi.h> as appropriate. Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Matthew Garrett <mjg59@srcf.ucam.org> Cc: Tony Luck <tony.luck@intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Acked-by: Bjorn Helgaas <bhelgaas@google.com> (drivers/pci stuff) Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> (Xen stuff) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-12-03 08:49:16 +08:00
#include <asm/io.h>
#include "internal.h"
#define ACPI_EC_CLASS "embedded_controller"
#define ACPI_EC_DEVICE_NAME "Embedded Controller"
#define ACPI_EC_FILE_INFO "info"
/* EC status register */
#define ACPI_EC_FLAG_OBF 0x01 /* Output buffer full */
#define ACPI_EC_FLAG_IBF 0x02 /* Input buffer full */
#define ACPI_EC_FLAG_CMD 0x08 /* Input buffer contains a command */
#define ACPI_EC_FLAG_BURST 0x10 /* burst mode */
#define ACPI_EC_FLAG_SCI 0x20 /* EC-SCI occurred */
/* EC commands */
enum ec_command {
ACPI_EC_COMMAND_READ = 0x80,
ACPI_EC_COMMAND_WRITE = 0x81,
ACPI_EC_BURST_ENABLE = 0x82,
ACPI_EC_BURST_DISABLE = 0x83,
ACPI_EC_COMMAND_QUERY = 0x84,
};
#define ACPI_EC_DELAY 500 /* Wait 500ms max. during EC ops */
#define ACPI_EC_UDELAY_GLK 1000 /* Wait 1ms max. to get global lock */
#define ACPI_EC_MSI_UDELAY 550 /* Wait 550us for MSI EC */
#define ACPI_EC_UDELAY_POLL 1000 /* Wait 1ms for EC transaction polling */
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
#define ACPI_EC_CLEAR_MAX 100 /* Maximum number of events to query
* when trying to clear the EC */
enum {
EC_FLAGS_QUERY_PENDING, /* Query is pending */
EC_FLAGS_GPE_STORM, /* GPE storm detected */
EC_FLAGS_HANDLERS_INSTALLED, /* Handlers for GPE and
* OpReg are installed */
EC_FLAGS_STARTED, /* Driver is started */
EC_FLAGS_STOPPED, /* Driver is stopped */
};
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
#define ACPI_EC_COMMAND_POLL 0x01 /* Available for command byte */
#define ACPI_EC_COMMAND_COMPLETE 0x02 /* Completed last byte */
/* ec.c is compiled in acpi namespace so this shows up as acpi.ec_delay param */
static unsigned int ec_delay __read_mostly = ACPI_EC_DELAY;
module_param(ec_delay, uint, 0644);
MODULE_PARM_DESC(ec_delay, "Timeout(ms) waited until an EC command completes");
/*
* If the number of false interrupts per one transaction exceeds
* this threshold, will think there is a GPE storm happened and
* will disable the GPE for normal transaction.
*/
static unsigned int ec_storm_threshold __read_mostly = 8;
module_param(ec_storm_threshold, uint, 0644);
MODULE_PARM_DESC(ec_storm_threshold, "Maxim false GPE numbers not considered as GPE storm");
struct acpi_ec_query_handler {
struct list_head node;
acpi_ec_query_func func;
acpi_handle handle;
void *data;
u8 query_bit;
struct kref kref;
};
struct transaction {
const u8 *wdata;
u8 *rdata;
unsigned short irq_count;
u8 command;
u8 wi;
u8 ri;
u8 wlen;
u8 rlen;
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
u8 flags;
unsigned long timestamp;
};
static int acpi_ec_query(struct acpi_ec *ec, u8 *data);
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
static void advance_transaction(struct acpi_ec *ec);
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
struct acpi_ec *boot_ec, *first_ec;
EXPORT_SYMBOL(first_ec);
static int EC_FLAGS_MSI; /* Out-of-spec MSI controller */
static int EC_FLAGS_VALIDATE_ECDT; /* ASUStec ECDTs need to be validated */
static int EC_FLAGS_SKIP_DSDT_SCAN; /* Not all BIOS survive early DSDT scan */
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
static int EC_FLAGS_CLEAR_ON_RESUME; /* Needs acpi_ec_clear() on boot/resume */
static int EC_FLAGS_QUERY_HANDSHAKE; /* Needs QR_EC issued when SCI_EVT set */
/* --------------------------------------------------------------------------
* Device Flags
* -------------------------------------------------------------------------- */
static bool acpi_ec_started(struct acpi_ec *ec)
{
return test_bit(EC_FLAGS_STARTED, &ec->flags) &&
!test_bit(EC_FLAGS_STOPPED, &ec->flags);
}
static bool acpi_ec_flushed(struct acpi_ec *ec)
{
return ec->reference_count == 1;
}
/* --------------------------------------------------------------------------
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
* EC Registers
* -------------------------------------------------------------------------- */
static inline u8 acpi_ec_read_status(struct acpi_ec *ec)
{
u8 x = inb(ec->command_addr);
pr_debug("EC_SC(R) = 0x%2.2x "
"SCI_EVT=%d BURST=%d CMD=%d IBF=%d OBF=%d\n",
x,
!!(x & ACPI_EC_FLAG_SCI),
!!(x & ACPI_EC_FLAG_BURST),
!!(x & ACPI_EC_FLAG_CMD),
!!(x & ACPI_EC_FLAG_IBF),
!!(x & ACPI_EC_FLAG_OBF));
return x;
}
static inline u8 acpi_ec_read_data(struct acpi_ec *ec)
{
u8 x = inb(ec->data_addr);
ec->curr->timestamp = jiffies;
pr_debug("EC_DATA(R) = 0x%2.2x\n", x);
return x;
}
static inline void acpi_ec_write_cmd(struct acpi_ec *ec, u8 command)
{
pr_debug("EC_SC(W) = 0x%2.2x\n", command);
outb(command, ec->command_addr);
ec->curr->timestamp = jiffies;
}
static inline void acpi_ec_write_data(struct acpi_ec *ec, u8 data)
{
pr_debug("EC_DATA(W) = 0x%2.2x\n", data);
outb(data, ec->data_addr);
ec->curr->timestamp = jiffies;
}
#ifdef DEBUG
static const char *acpi_ec_cmd_string(u8 cmd)
{
switch (cmd) {
case 0x80:
return "RD_EC";
case 0x81:
return "WR_EC";
case 0x82:
return "BE_EC";
case 0x83:
return "BD_EC";
case 0x84:
return "QR_EC";
}
return "UNKNOWN";
}
#else
#define acpi_ec_cmd_string(cmd) "UNDEF"
#endif
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
/* --------------------------------------------------------------------------
* GPE Registers
* -------------------------------------------------------------------------- */
static inline bool acpi_ec_is_gpe_raised(struct acpi_ec *ec)
{
acpi_event_status gpe_status = 0;
(void)acpi_get_gpe_status(NULL, ec->gpe, &gpe_status);
return (gpe_status & ACPI_EVENT_FLAG_SET) ? true : false;
}
static inline void acpi_ec_enable_gpe(struct acpi_ec *ec, bool open)
{
if (open)
acpi_enable_gpe(NULL, ec->gpe);
else
acpi_set_gpe(NULL, ec->gpe, ACPI_GPE_ENABLE);
if (acpi_ec_is_gpe_raised(ec)) {
/*
* On some platforms, EN=1 writes cannot trigger GPE. So
* software need to manually trigger a pseudo GPE event on
* EN=1 writes.
*/
pr_debug("***** Polling quirk *****\n");
advance_transaction(ec);
}
}
static inline void acpi_ec_disable_gpe(struct acpi_ec *ec, bool close)
{
if (close)
acpi_disable_gpe(NULL, ec->gpe);
else
acpi_set_gpe(NULL, ec->gpe, ACPI_GPE_DISABLE);
}
static inline void acpi_ec_clear_gpe(struct acpi_ec *ec)
{
/*
* GPE STS is a W1C register, which means:
* 1. Software can clear it without worrying about clearing other
* GPEs' STS bits when the hardware sets them in parallel.
* 2. As long as software can ensure only clearing it when it is
* set, hardware won't set it in parallel.
* So software can clear GPE in any contexts.
* Warning: do not move the check into advance_transaction() as the
* EC commands will be sent without GPE raised.
*/
if (!acpi_ec_is_gpe_raised(ec))
return;
acpi_clear_gpe(NULL, ec->gpe);
}
/* --------------------------------------------------------------------------
* Transaction Management
* -------------------------------------------------------------------------- */
static void acpi_ec_submit_request(struct acpi_ec *ec)
{
ec->reference_count++;
if (ec->reference_count == 1)
acpi_ec_enable_gpe(ec, true);
}
static void acpi_ec_complete_request(struct acpi_ec *ec)
{
bool flushed = false;
ec->reference_count--;
if (ec->reference_count == 0)
acpi_ec_disable_gpe(ec, true);
flushed = acpi_ec_flushed(ec);
if (flushed)
wake_up(&ec->wait);
}
/*
* acpi_ec_submit_flushable_request() - Increase the reference count unless
* the flush operation is not in
* progress
* @ec: the EC device
*
* This function must be used before taking a new action that should hold
* the reference count. If this function returns false, then the action
* must be discarded or it will prevent the flush operation from being
* completed.
*/
static bool acpi_ec_submit_flushable_request(struct acpi_ec *ec)
{
if (!acpi_ec_started(ec))
return false;
acpi_ec_submit_request(ec);
return true;
}
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
static void acpi_ec_submit_query(struct acpi_ec *ec)
{
if (!test_and_set_bit(EC_FLAGS_QUERY_PENDING, &ec->flags)) {
pr_debug("***** Event started *****\n");
schedule_work(&ec->work);
}
}
static void acpi_ec_complete_query(struct acpi_ec *ec)
{
if (ec->curr->command == ACPI_EC_COMMAND_QUERY) {
clear_bit(EC_FLAGS_QUERY_PENDING, &ec->flags);
pr_debug("***** Event stopped *****\n");
}
}
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
static int ec_transaction_completed(struct acpi_ec *ec)
{
unsigned long flags;
int ret = 0;
spin_lock_irqsave(&ec->lock, flags);
if (ec->curr && (ec->curr->flags & ACPI_EC_COMMAND_COMPLETE))
ret = 1;
spin_unlock_irqrestore(&ec->lock, flags);
return ret;
}
static void advance_transaction(struct acpi_ec *ec)
{
struct transaction *t;
ACPI / EC: Avoid race condition related to advance_transaction() The advance_transaction() will be invoked from the IRQ context GPE handler and the task context ec_poll(). The handling of this function is locked so that the EC state machine are ensured to be advanced sequentially. But there is a problem. Before invoking advance_transaction(), EC_SC(R) is read. Then for advance_transaction(), there could be race condition around the lock from both contexts. The first one reading the register could fail this race and when it passes the stale register value to the state machine advancement code, the hardware condition is totally different from when the register is read. And the hardware accesses determined from the wrong hardware status can break the EC state machine. And there could be cases that the functionalities of the platform firmware are seriously affected. For example: 1. When 2 EC_DATA(W) writes compete the IBF=0, the 2nd EC_DATA(W) write may be invalid due to IBF=1 after the 1st EC_DATA(W) write. Then the hardware will either refuse to respond a next EC_SC(W) write of the next command or discard the current WR_EC command when it receives a EC_SC(W) write of the next command. 2. When 1 EC_SC(W) write and 1 EC_DATA(W) write compete the IBF=0, the EC_DATA(W) write may be invalid due to IBF=1 after the EC_SC(W) write. The next EC_DATA(R) could never be responded by the hardware. This is the root cause of the reported issue. Fix this issue by moving the EC_SC(R) access into the lock so that we can ensure that the state machine is advanced consistently. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:17 +08:00
u8 status;
bool wakeup = false;
pr_debug("===== %s (%d) =====\n",
in_interrupt() ? "IRQ" : "TASK", smp_processor_id());
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
/*
* By always clearing STS before handling all indications, we can
* ensure a hardware STS 0->1 change after this clearing can always
* trigger a GPE interrupt.
*/
acpi_ec_clear_gpe(ec);
ACPI / EC: Avoid race condition related to advance_transaction() The advance_transaction() will be invoked from the IRQ context GPE handler and the task context ec_poll(). The handling of this function is locked so that the EC state machine are ensured to be advanced sequentially. But there is a problem. Before invoking advance_transaction(), EC_SC(R) is read. Then for advance_transaction(), there could be race condition around the lock from both contexts. The first one reading the register could fail this race and when it passes the stale register value to the state machine advancement code, the hardware condition is totally different from when the register is read. And the hardware accesses determined from the wrong hardware status can break the EC state machine. And there could be cases that the functionalities of the platform firmware are seriously affected. For example: 1. When 2 EC_DATA(W) writes compete the IBF=0, the 2nd EC_DATA(W) write may be invalid due to IBF=1 after the 1st EC_DATA(W) write. Then the hardware will either refuse to respond a next EC_SC(W) write of the next command or discard the current WR_EC command when it receives a EC_SC(W) write of the next command. 2. When 1 EC_SC(W) write and 1 EC_DATA(W) write compete the IBF=0, the EC_DATA(W) write may be invalid due to IBF=1 after the EC_SC(W) write. The next EC_DATA(R) could never be responded by the hardware. This is the root cause of the reported issue. Fix this issue by moving the EC_SC(R) access into the lock so that we can ensure that the state machine is advanced consistently. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:17 +08:00
status = acpi_ec_read_status(ec);
t = ec->curr;
if (!t)
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
goto err;
if (t->flags & ACPI_EC_COMMAND_POLL) {
if (t->wlen > t->wi) {
if ((status & ACPI_EC_FLAG_IBF) == 0)
acpi_ec_write_data(ec, t->wdata[t->wi++]);
else
goto err;
} else if (t->rlen > t->ri) {
if ((status & ACPI_EC_FLAG_OBF) == 1) {
t->rdata[t->ri++] = acpi_ec_read_data(ec);
if (t->rlen == t->ri) {
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
t->flags |= ACPI_EC_COMMAND_COMPLETE;
ACPI / EC: Add support to disallow QR_EC to be issued when SCI_EVT isn't set There is a platform refusing to respond QR_EC when SCI_EVT isn't set (Acer Aspire V5-573G). Currently, we rely on the behaviour that the EC firmware can respond something (for example, 0x00 to indicate "no outstanding events") to QR_EC even when SCI_EVT is not set, but the reporter has complained about AC/battery pluging/unpluging and video brightness change delay on that platform. This is because the work item that has issued QR_EC has to wait until timeout in this case, and the _Qxx method evaluation work item queued after QR_EC one is delayed. It sounds reasonable to fix this issue by: 1. Implementing SCI_EVT sanity check before issuing QR_EC in the EC driver's main state machine. 2. Moving QR_EC issuing out of the work queue used by _Qxx evaluation to a seperate IRQ handling thread. This patch fixes this issue using solution 1. By disallowing QR_EC to be issued when SCI_EVT isn't set, we are able to handle such platform in the EC driver's main state machine. This patch enhances the state machine in this way to survive with such malfunctioning EC firmware. Note that this patch can also fix CLEAR_ON_RESUME quirk which also relies on the assumption that the platforms are able to respond even when SCI_EVT isn't set. Fixes: c0d653412fc8 ACPI / EC: Fix race condition in ec_transaction_completed() Link: https://bugzilla.kernel.org/show_bug.cgi?id=82611 Reported-and-tested-by: Alexander Mezin <mezin.alexander@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: 3.16+ <stable@vger.kernel.org> # 3.16+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-08-21 14:41:13 +08:00
if (t->command == ACPI_EC_COMMAND_QUERY)
pr_debug("***** Command(%s) hardware completion *****\n",
acpi_ec_cmd_string(t->command));
wakeup = true;
}
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
} else
goto err;
} else if (t->wlen == t->wi &&
(status & ACPI_EC_FLAG_IBF) == 0) {
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
t->flags |= ACPI_EC_COMMAND_COMPLETE;
wakeup = true;
}
goto out;
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
} else {
if (EC_FLAGS_QUERY_HANDSHAKE &&
!(status & ACPI_EC_FLAG_SCI) &&
ACPI / EC: Add support to disallow QR_EC to be issued when SCI_EVT isn't set There is a platform refusing to respond QR_EC when SCI_EVT isn't set (Acer Aspire V5-573G). Currently, we rely on the behaviour that the EC firmware can respond something (for example, 0x00 to indicate "no outstanding events") to QR_EC even when SCI_EVT is not set, but the reporter has complained about AC/battery pluging/unpluging and video brightness change delay on that platform. This is because the work item that has issued QR_EC has to wait until timeout in this case, and the _Qxx method evaluation work item queued after QR_EC one is delayed. It sounds reasonable to fix this issue by: 1. Implementing SCI_EVT sanity check before issuing QR_EC in the EC driver's main state machine. 2. Moving QR_EC issuing out of the work queue used by _Qxx evaluation to a seperate IRQ handling thread. This patch fixes this issue using solution 1. By disallowing QR_EC to be issued when SCI_EVT isn't set, we are able to handle such platform in the EC driver's main state machine. This patch enhances the state machine in this way to survive with such malfunctioning EC firmware. Note that this patch can also fix CLEAR_ON_RESUME quirk which also relies on the assumption that the platforms are able to respond even when SCI_EVT isn't set. Fixes: c0d653412fc8 ACPI / EC: Fix race condition in ec_transaction_completed() Link: https://bugzilla.kernel.org/show_bug.cgi?id=82611 Reported-and-tested-by: Alexander Mezin <mezin.alexander@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: 3.16+ <stable@vger.kernel.org> # 3.16+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-08-21 14:41:13 +08:00
(t->command == ACPI_EC_COMMAND_QUERY)) {
t->flags |= ACPI_EC_COMMAND_POLL;
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
acpi_ec_complete_query(ec);
ACPI / EC: Add support to disallow QR_EC to be issued when SCI_EVT isn't set There is a platform refusing to respond QR_EC when SCI_EVT isn't set (Acer Aspire V5-573G). Currently, we rely on the behaviour that the EC firmware can respond something (for example, 0x00 to indicate "no outstanding events") to QR_EC even when SCI_EVT is not set, but the reporter has complained about AC/battery pluging/unpluging and video brightness change delay on that platform. This is because the work item that has issued QR_EC has to wait until timeout in this case, and the _Qxx method evaluation work item queued after QR_EC one is delayed. It sounds reasonable to fix this issue by: 1. Implementing SCI_EVT sanity check before issuing QR_EC in the EC driver's main state machine. 2. Moving QR_EC issuing out of the work queue used by _Qxx evaluation to a seperate IRQ handling thread. This patch fixes this issue using solution 1. By disallowing QR_EC to be issued when SCI_EVT isn't set, we are able to handle such platform in the EC driver's main state machine. This patch enhances the state machine in this way to survive with such malfunctioning EC firmware. Note that this patch can also fix CLEAR_ON_RESUME quirk which also relies on the assumption that the platforms are able to respond even when SCI_EVT isn't set. Fixes: c0d653412fc8 ACPI / EC: Fix race condition in ec_transaction_completed() Link: https://bugzilla.kernel.org/show_bug.cgi?id=82611 Reported-and-tested-by: Alexander Mezin <mezin.alexander@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: 3.16+ <stable@vger.kernel.org> # 3.16+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-08-21 14:41:13 +08:00
t->rdata[t->ri++] = 0x00;
t->flags |= ACPI_EC_COMMAND_COMPLETE;
pr_debug("***** Command(%s) software completion *****\n",
acpi_ec_cmd_string(t->command));
ACPI / EC: Add support to disallow QR_EC to be issued when SCI_EVT isn't set There is a platform refusing to respond QR_EC when SCI_EVT isn't set (Acer Aspire V5-573G). Currently, we rely on the behaviour that the EC firmware can respond something (for example, 0x00 to indicate "no outstanding events") to QR_EC even when SCI_EVT is not set, but the reporter has complained about AC/battery pluging/unpluging and video brightness change delay on that platform. This is because the work item that has issued QR_EC has to wait until timeout in this case, and the _Qxx method evaluation work item queued after QR_EC one is delayed. It sounds reasonable to fix this issue by: 1. Implementing SCI_EVT sanity check before issuing QR_EC in the EC driver's main state machine. 2. Moving QR_EC issuing out of the work queue used by _Qxx evaluation to a seperate IRQ handling thread. This patch fixes this issue using solution 1. By disallowing QR_EC to be issued when SCI_EVT isn't set, we are able to handle such platform in the EC driver's main state machine. This patch enhances the state machine in this way to survive with such malfunctioning EC firmware. Note that this patch can also fix CLEAR_ON_RESUME quirk which also relies on the assumption that the platforms are able to respond even when SCI_EVT isn't set. Fixes: c0d653412fc8 ACPI / EC: Fix race condition in ec_transaction_completed() Link: https://bugzilla.kernel.org/show_bug.cgi?id=82611 Reported-and-tested-by: Alexander Mezin <mezin.alexander@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: 3.16+ <stable@vger.kernel.org> # 3.16+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-08-21 14:41:13 +08:00
wakeup = true;
} else if ((status & ACPI_EC_FLAG_IBF) == 0) {
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
acpi_ec_write_cmd(ec, t->command);
t->flags |= ACPI_EC_COMMAND_POLL;
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
acpi_ec_complete_query(ec);
} else
goto err;
goto out;
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
}
err:
/*
* If SCI bit is set, then don't think it's a false IRQ
* otherwise will take a not handled IRQ as a false one.
*/
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
if (!(status & ACPI_EC_FLAG_SCI)) {
if (in_interrupt() && t)
++t->irq_count;
}
out:
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
if (status & ACPI_EC_FLAG_SCI)
acpi_ec_submit_query(ec);
if (wakeup && in_interrupt())
wake_up(&ec->wait);
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
}
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
static void start_transaction(struct acpi_ec *ec)
{
ec->curr->irq_count = ec->curr->wi = ec->curr->ri = 0;
ec->curr->flags = 0;
ec->curr->timestamp = jiffies;
advance_transaction(ec);
}
static int ec_poll(struct acpi_ec *ec)
{
unsigned long flags;
int repeat = 5; /* number of command restarts */
while (repeat--) {
unsigned long delay = jiffies +
msecs_to_jiffies(ec_delay);
unsigned long usecs = ACPI_EC_UDELAY_POLL;
do {
/* don't sleep with disabled interrupts */
if (EC_FLAGS_MSI || irqs_disabled()) {
usecs = ACPI_EC_MSI_UDELAY;
udelay(usecs);
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
if (ec_transaction_completed(ec))
return 0;
} else {
if (wait_event_timeout(ec->wait,
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
ec_transaction_completed(ec),
usecs_to_jiffies(usecs)))
return 0;
}
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
spin_lock_irqsave(&ec->lock, flags);
if (time_after(jiffies,
ec->curr->timestamp +
usecs_to_jiffies(usecs)))
advance_transaction(ec);
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
spin_unlock_irqrestore(&ec->lock, flags);
} while (time_before(jiffies, delay));
pr_debug("controller reset, restart transaction\n");
spin_lock_irqsave(&ec->lock, flags);
start_transaction(ec);
spin_unlock_irqrestore(&ec->lock, flags);
}
return -ETIME;
}
static int acpi_ec_transaction_unlocked(struct acpi_ec *ec,
struct transaction *t)
{
unsigned long tmp;
int ret = 0;
if (EC_FLAGS_MSI)
udelay(ACPI_EC_MSI_UDELAY);
/* start transaction */
spin_lock_irqsave(&ec->lock, tmp);
/* Enable GPE for command processing (IBF=0/OBF=1) */
if (!acpi_ec_submit_flushable_request(ec)) {
ret = -EINVAL;
goto unlock;
}
/* following two actions should be kept atomic */
ec->curr = t;
pr_debug("***** Command(%s) started *****\n",
acpi_ec_cmd_string(t->command));
start_transaction(ec);
Revert "ACPI / EC: Add support to disallow QR_EC to be issued before completing previous QR_EC" It is reported that the following commit breaks Samsung hardware: Commit: 558e4736f2e1b0e6323adf7a5e4df77ed6cfc1a4. Subject: ACPI / EC: Add support to disallow QR_EC to be issued before completing previous QR_EC Which means the Samsung behavior conflicts with the Acer behavior. 1. Samsung may behave like: [ +event 1 ] SCI_EVT set [ +event 2 ] SCI_EVT set write QR_EC read event [ -event 1 ] SCI_EVT clear Without the above commit, Samsung can work: [ +event 1 ] SCI_EVT set [ +event 2 ] SCI_EVT set write QR_EC CAN prepare next QR_EC as SCI_EVT=1 read event [ -event 1 ] SCI_EVT clear write QR_EC read event [ -event 2 ] SCI_EVT clear With the above commit, Samsung cannot work: [ +event 1 ] SCI_EVT set [ +event 2 ] SCI_EVT set write QR_EC read event [ -event 1 ] SCI_EVT clear CANNOT prepare next QR_EC as SCI_EVT=0 2. Acer may behave like: [ +event 1 ] SCI_EVT set [ +event 2 ] write QR_EC read event [ -event 1 ] SCI_EVT clear [ +event 2 ] SCI_EVT set Without the above commit, Acer cannot work when there is only 1 event: [ +event 1 ] SCI_EVT set write QR_EC can prepared next QR_EC as SCI_EVT=1 read event [ -event 1 ] SCI_EVT clear CANNOT write QR_EC as SCI_EVT=0 With the above commit, Acer can work: [ +event 1 ] SCI_EVT set [ +event 2 ] write QR_EC read event [ -event 1 ] SCI_EVT set can prepare next QR_EC because SCI_EVT=0 CAN write QR_EC as SCI_EVT=1 Since Acer can also work with only the following commit applied: Commit: 3afcf2ece453e1a8c2c6de19cdf06da3772a1b08 Subject: ACPI / EC: Add support to disallow QR_EC to be issued when SCI_EVT isn't set commit 558e4736f2e1b0e6323adf7a5e4df77ed6cfc1a4 can be reverted. Fixes: 558e4736f2e1 (ACPI / EC: Add support to disallow QR_EC to be issued ...) Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-and-tested-by: Ortwin Glück <odi@odi.ch> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: 3.17+ <stable@vger.kernel.org> # 3.17+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-10-29 11:33:43 +08:00
spin_unlock_irqrestore(&ec->lock, tmp);
ret = ec_poll(ec);
spin_lock_irqsave(&ec->lock, tmp);
pr_debug("***** Command(%s) stopped *****\n",
acpi_ec_cmd_string(t->command));
ec->curr = NULL;
/* Disable GPE for command processing (IBF=0/OBF=1) */
acpi_ec_complete_request(ec);
unlock:
spin_unlock_irqrestore(&ec->lock, tmp);
return ret;
}
static int acpi_ec_transaction(struct acpi_ec *ec, struct transaction *t)
{
int status;
u32 glk;
if (!ec || (!t) || (t->wlen && !t->wdata) || (t->rlen && !t->rdata))
return -EINVAL;
if (t->rdata)
memset(t->rdata, 0, t->rlen);
mutex_lock(&ec->mutex);
if (ec->global_lock) {
status = acpi_acquire_global_lock(ACPI_EC_UDELAY_GLK, &glk);
if (ACPI_FAILURE(status)) {
status = -ENODEV;
goto unlock;
}
}
/* disable GPE during transaction if storm is detected */
if (test_bit(EC_FLAGS_GPE_STORM, &ec->flags)) {
/* It has to be disabled, so that it doesn't trigger. */
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
acpi_ec_disable_gpe(ec, false);
}
status = acpi_ec_transaction_unlocked(ec, t);
if (test_bit(EC_FLAGS_GPE_STORM, &ec->flags)) {
msleep(1);
/* It is safe to enable the GPE outside of the transaction. */
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
acpi_ec_enable_gpe(ec, false);
} else if (t->irq_count > ec_storm_threshold) {
pr_info("GPE storm detected(%d GPEs), "
"transactions will use polling mode\n",
t->irq_count);
set_bit(EC_FLAGS_GPE_STORM, &ec->flags);
}
if (ec->global_lock)
acpi_release_global_lock(glk);
unlock:
mutex_unlock(&ec->mutex);
return status;
}
static int acpi_ec_burst_enable(struct acpi_ec *ec)
{
u8 d;
struct transaction t = {.command = ACPI_EC_BURST_ENABLE,
.wdata = NULL, .rdata = &d,
.wlen = 0, .rlen = 1};
return acpi_ec_transaction(ec, &t);
}
static int acpi_ec_burst_disable(struct acpi_ec *ec)
{
struct transaction t = {.command = ACPI_EC_BURST_DISABLE,
.wdata = NULL, .rdata = NULL,
.wlen = 0, .rlen = 0};
return (acpi_ec_read_status(ec) & ACPI_EC_FLAG_BURST) ?
acpi_ec_transaction(ec, &t) : 0;
}
static int acpi_ec_read(struct acpi_ec *ec, u8 address, u8 *data)
{
int result;
u8 d;
struct transaction t = {.command = ACPI_EC_COMMAND_READ,
.wdata = &address, .rdata = &d,
.wlen = 1, .rlen = 1};
result = acpi_ec_transaction(ec, &t);
*data = d;
return result;
}
static int acpi_ec_write(struct acpi_ec *ec, u8 address, u8 data)
{
u8 wdata[2] = { address, data };
struct transaction t = {.command = ACPI_EC_COMMAND_WRITE,
.wdata = wdata, .rdata = NULL,
.wlen = 2, .rlen = 0};
return acpi_ec_transaction(ec, &t);
}
int ec_read(u8 addr, u8 *val)
{
int err;
u8 temp_data;
if (!first_ec)
return -ENODEV;
err = acpi_ec_read(first_ec, addr, &temp_data);
if (!err) {
*val = temp_data;
return 0;
}
return err;
}
EXPORT_SYMBOL(ec_read);
int ec_write(u8 addr, u8 val)
{
int err;
if (!first_ec)
return -ENODEV;
err = acpi_ec_write(first_ec, addr, val);
return err;
}
EXPORT_SYMBOL(ec_write);
int ec_transaction(u8 command,
const u8 *wdata, unsigned wdata_len,
u8 *rdata, unsigned rdata_len)
{
struct transaction t = {.command = command,
.wdata = wdata, .rdata = rdata,
.wlen = wdata_len, .rlen = rdata_len};
if (!first_ec)
return -ENODEV;
return acpi_ec_transaction(first_ec, &t);
}
EXPORT_SYMBOL(ec_transaction);
/* Get the handle to the EC device */
acpi_handle ec_get_handle(void)
{
if (!first_ec)
return NULL;
return first_ec->handle;
}
EXPORT_SYMBOL(ec_get_handle);
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
/*
ACPI / EC: Process rather than discard events in acpi_ec_clear Address a regression caused by commit ad332c8a4533: (ACPI / EC: Clear stale EC events on Samsung systems) After the earlier patch, there was found to be a race condition on some earlier Samsung systems (N150/N210/N220). The function acpi_ec_clear was sometimes discarding a new EC event before its GPE was triggered by the system. In the case of these systems, this meant that the "lid open" event was not registered on resume if that was the cause of the wake, leading to problems when attempting to close the lid to suspend again. After testing on a number of Samsung systems, both those affected by the previous EC bug and those affected by the race condition, it seemed that the best course of action was to process rather than discard the events. On Samsung systems which accumulate stale EC events, there does not seem to be any adverse side-effects of running the associated _Q methods. This patch adds an argument to the static function acpi_ec_sync_query so that it may be used within the acpi_ec_clear loop in place of acpi_ec_query_unlocked which was used previously. With thanks to Stefan Biereigel for reporting the issue, and for all the people who helped test the new patch on affected systems. Fixes: ad332c8a4533 (ACPI / EC: Clear stale EC events on Samsung systems) References: https://lkml.kernel.org/r/532FE3B2.9060808@biereigel-wb.de References: https://bugzilla.kernel.org/show_bug.cgi?id=44161#c173 Reported-by: Stefan Biereigel <stefan@biereigel.de> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Stefan Biereigel <stefan@biereigel.de> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Nicolas Porcel <nicolasporcel06@gmail.com> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Giannis Koutsou <giannis.koutsou@gmail.com> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Cc: 3.14+ <stable@vger.kernel.org> # 3.14+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-04-29 22:51:20 +08:00
* Process _Q events that might have accumulated in the EC.
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
* Run with locked ec mutex.
*/
static void acpi_ec_clear(struct acpi_ec *ec)
{
int i, status;
u8 value = 0;
for (i = 0; i < ACPI_EC_CLEAR_MAX; i++) {
status = acpi_ec_query(ec, &value);
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
if (status || !value)
break;
}
if (unlikely(i == ACPI_EC_CLEAR_MAX))
pr_warn("Warning: Maximum of %d stale EC events cleared\n", i);
else
pr_info("%d stale EC events cleared\n", i);
}
static void acpi_ec_start(struct acpi_ec *ec, bool resuming)
{
unsigned long flags;
spin_lock_irqsave(&ec->lock, flags);
if (!test_and_set_bit(EC_FLAGS_STARTED, &ec->flags)) {
pr_debug("+++++ Starting EC +++++\n");
/* Enable GPE for event processing (SCI_EVT=1) */
if (!resuming)
acpi_ec_submit_request(ec);
pr_info("+++++ EC started +++++\n");
}
spin_unlock_irqrestore(&ec->lock, flags);
}
static bool acpi_ec_stopped(struct acpi_ec *ec)
{
unsigned long flags;
bool flushed;
spin_lock_irqsave(&ec->lock, flags);
flushed = acpi_ec_flushed(ec);
spin_unlock_irqrestore(&ec->lock, flags);
return flushed;
}
static void acpi_ec_stop(struct acpi_ec *ec, bool suspending)
{
unsigned long flags;
spin_lock_irqsave(&ec->lock, flags);
if (acpi_ec_started(ec)) {
pr_debug("+++++ Stopping EC +++++\n");
set_bit(EC_FLAGS_STOPPED, &ec->flags);
spin_unlock_irqrestore(&ec->lock, flags);
wait_event(ec->wait, acpi_ec_stopped(ec));
spin_lock_irqsave(&ec->lock, flags);
/* Disable GPE for event processing (SCI_EVT=1) */
if (!suspending)
acpi_ec_complete_request(ec);
clear_bit(EC_FLAGS_STARTED, &ec->flags);
clear_bit(EC_FLAGS_STOPPED, &ec->flags);
pr_info("+++++ EC stopped +++++\n");
}
spin_unlock_irqrestore(&ec->lock, flags);
}
void acpi_ec_block_transactions(void)
{
struct acpi_ec *ec = first_ec;
if (!ec)
return;
mutex_lock(&ec->mutex);
/* Prevent transactions from being carried out */
acpi_ec_stop(ec, true);
mutex_unlock(&ec->mutex);
}
void acpi_ec_unblock_transactions(void)
{
struct acpi_ec *ec = first_ec;
if (!ec)
return;
/* Allow transactions to be carried out again */
acpi_ec_start(ec, true);
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
if (EC_FLAGS_CLEAR_ON_RESUME)
acpi_ec_clear(ec);
}
void acpi_ec_unblock_transactions_early(void)
{
/*
* Allow transactions to happen again (this function is called from
* atomic context during wakeup, so we don't need to acquire the mutex).
*/
if (first_ec)
acpi_ec_start(first_ec, true);
}
/* --------------------------------------------------------------------------
Event Management
-------------------------------------------------------------------------- */
static struct acpi_ec_query_handler *
acpi_ec_get_query_handler(struct acpi_ec_query_handler *handler)
{
if (handler)
kref_get(&handler->kref);
return handler;
}
static void acpi_ec_query_handler_release(struct kref *kref)
{
struct acpi_ec_query_handler *handler =
container_of(kref, struct acpi_ec_query_handler, kref);
kfree(handler);
}
static void acpi_ec_put_query_handler(struct acpi_ec_query_handler *handler)
{
kref_put(&handler->kref, acpi_ec_query_handler_release);
}
int acpi_ec_add_query_handler(struct acpi_ec *ec, u8 query_bit,
acpi_handle handle, acpi_ec_query_func func,
void *data)
{
struct acpi_ec_query_handler *handler =
kzalloc(sizeof(struct acpi_ec_query_handler), GFP_KERNEL);
if (!handler)
return -ENOMEM;
handler->query_bit = query_bit;
handler->handle = handle;
handler->func = func;
handler->data = data;
mutex_lock(&ec->mutex);
kref_init(&handler->kref);
list_add(&handler->node, &ec->list);
mutex_unlock(&ec->mutex);
return 0;
}
EXPORT_SYMBOL_GPL(acpi_ec_add_query_handler);
void acpi_ec_remove_query_handler(struct acpi_ec *ec, u8 query_bit)
{
struct acpi_ec_query_handler *handler, *tmp;
LIST_HEAD(free_list);
mutex_lock(&ec->mutex);
list_for_each_entry_safe(handler, tmp, &ec->list, node) {
if (query_bit == handler->query_bit) {
list_del_init(&handler->node);
list_add(&handler->node, &free_list);
}
}
mutex_unlock(&ec->mutex);
list_for_each_entry(handler, &free_list, node)
acpi_ec_put_query_handler(handler);
}
EXPORT_SYMBOL_GPL(acpi_ec_remove_query_handler);
static void acpi_ec_run(void *cxt)
{
struct acpi_ec_query_handler *handler = cxt;
if (!handler)
return;
pr_debug("##### Query(0x%02x) started #####\n", handler->query_bit);
if (handler->func)
handler->func(handler->data);
else if (handler->handle)
acpi_evaluate_object(handler->handle, NULL, NULL, NULL);
pr_debug("##### Query(0x%02x) stopped #####\n", handler->query_bit);
acpi_ec_put_query_handler(handler);
}
static int acpi_ec_query(struct acpi_ec *ec, u8 *data)
{
u8 value = 0;
int result;
acpi_status status;
struct acpi_ec_query_handler *handler;
struct transaction t = {.command = ACPI_EC_COMMAND_QUERY,
.wdata = NULL, .rdata = &value,
.wlen = 0, .rlen = 1};
ACPI / EC: Process rather than discard events in acpi_ec_clear Address a regression caused by commit ad332c8a4533: (ACPI / EC: Clear stale EC events on Samsung systems) After the earlier patch, there was found to be a race condition on some earlier Samsung systems (N150/N210/N220). The function acpi_ec_clear was sometimes discarding a new EC event before its GPE was triggered by the system. In the case of these systems, this meant that the "lid open" event was not registered on resume if that was the cause of the wake, leading to problems when attempting to close the lid to suspend again. After testing on a number of Samsung systems, both those affected by the previous EC bug and those affected by the race condition, it seemed that the best course of action was to process rather than discard the events. On Samsung systems which accumulate stale EC events, there does not seem to be any adverse side-effects of running the associated _Q methods. This patch adds an argument to the static function acpi_ec_sync_query so that it may be used within the acpi_ec_clear loop in place of acpi_ec_query_unlocked which was used previously. With thanks to Stefan Biereigel for reporting the issue, and for all the people who helped test the new patch on affected systems. Fixes: ad332c8a4533 (ACPI / EC: Clear stale EC events on Samsung systems) References: https://lkml.kernel.org/r/532FE3B2.9060808@biereigel-wb.de References: https://bugzilla.kernel.org/show_bug.cgi?id=44161#c173 Reported-by: Stefan Biereigel <stefan@biereigel.de> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Stefan Biereigel <stefan@biereigel.de> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Nicolas Porcel <nicolasporcel06@gmail.com> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Giannis Koutsou <giannis.koutsou@gmail.com> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Cc: 3.14+ <stable@vger.kernel.org> # 3.14+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-04-29 22:51:20 +08:00
/*
* Query the EC to find out which _Qxx method we need to evaluate.
* Note that successful completion of the query causes the ACPI_EC_SCI
* bit to be cleared (and thus clearing the interrupt source).
*/
result = acpi_ec_transaction(ec, &t);
if (result)
return result;
if (data)
*data = value;
if (!value)
return -ENODATA;
ACPI / EC: Process rather than discard events in acpi_ec_clear Address a regression caused by commit ad332c8a4533: (ACPI / EC: Clear stale EC events on Samsung systems) After the earlier patch, there was found to be a race condition on some earlier Samsung systems (N150/N210/N220). The function acpi_ec_clear was sometimes discarding a new EC event before its GPE was triggered by the system. In the case of these systems, this meant that the "lid open" event was not registered on resume if that was the cause of the wake, leading to problems when attempting to close the lid to suspend again. After testing on a number of Samsung systems, both those affected by the previous EC bug and those affected by the race condition, it seemed that the best course of action was to process rather than discard the events. On Samsung systems which accumulate stale EC events, there does not seem to be any adverse side-effects of running the associated _Q methods. This patch adds an argument to the static function acpi_ec_sync_query so that it may be used within the acpi_ec_clear loop in place of acpi_ec_query_unlocked which was used previously. With thanks to Stefan Biereigel for reporting the issue, and for all the people who helped test the new patch on affected systems. Fixes: ad332c8a4533 (ACPI / EC: Clear stale EC events on Samsung systems) References: https://lkml.kernel.org/r/532FE3B2.9060808@biereigel-wb.de References: https://bugzilla.kernel.org/show_bug.cgi?id=44161#c173 Reported-by: Stefan Biereigel <stefan@biereigel.de> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Stefan Biereigel <stefan@biereigel.de> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Nicolas Porcel <nicolasporcel06@gmail.com> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Giannis Koutsou <giannis.koutsou@gmail.com> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Cc: 3.14+ <stable@vger.kernel.org> # 3.14+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-04-29 22:51:20 +08:00
mutex_lock(&ec->mutex);
list_for_each_entry(handler, &ec->list, node) {
if (value == handler->query_bit) {
/* have custom handler for this bit */
handler = acpi_ec_get_query_handler(handler);
pr_debug("##### Query(0x%02x) scheduled #####\n",
handler->query_bit);
status = acpi_os_execute((handler->func) ?
OSL_NOTIFY_HANDLER : OSL_GPE_HANDLER,
acpi_ec_run, handler);
if (ACPI_FAILURE(status))
result = -EBUSY;
break;
}
}
mutex_unlock(&ec->mutex);
return result;
}
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
static void acpi_ec_gpe_poller(struct work_struct *work)
{
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
struct acpi_ec *ec = container_of(work, struct acpi_ec, work);
acpi_ec_query(ec, NULL);
}
static u32 acpi_ec_gpe_handler(acpi_handle gpe_device,
u32 gpe_number, void *data)
{
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
unsigned long flags;
struct acpi_ec *ec = data;
ACPI / EC: Add asynchronous command byte write support Move the first command byte write into advance_transaction() so that all EC register accesses that can affect the command processing state machine can happen in this asynchronous state machine advancement function. The advance_transaction() function then can be a complete implementation of an asyncrhonous transaction for a single command so that: 1. The first command byte can be written in the interrupt context; 2. The command completion waiter can also be used to wait the first command byte's timeout; 3. In BURST mode, the follow-up command bytes can be written in the interrupt context directly, so that it doesn't need to return to the task context. Returning to the task context reduces the throughput of the BURST mode and in the worst cases where the system workload is very high, this leads to the hardware driven automatic BURST mode exit. In order not to increase memory consumption, convert 'done' into 'flags' to contain multiple indications: 1. ACPI_EC_COMMAND_COMPLETE: converting from original 'done' condition, indicating the completion of the command transaction. 2. ACPI_EC_COMMAND_POLL: indicating the availability of writing the first command byte. A new command can utilize this flag to compete for the right of accessing the underlying hardware. There is a follow-up bug fix that has utilized this new flag. The 2 flags are important because it also reflects a key concept of IO programs' design used in the system softwares. Normally an IO program running in the kernel should first be implemented in the asynchronous way. And the 2 flags are the most common way to implement its synchronous operations on top of the asynchronous operations: 1. POLL: This flag can be used to block until the asynchronous operations can happen. 2. COMPLETE: This flag can be used to block until the asynchronous operations have completed. By constructing code cleanly in this way, many difficult problems can be solved smoothly. Link: https://bugzilla.kernel.org/show_bug.cgi?id=70891 Link: https://bugzilla.kernel.org/show_bug.cgi?id=63931 Link: https://bugzilla.kernel.org/show_bug.cgi?id=59911 Reported-and-tested-by: Gareth Williams <gareth@garethwilliams.me.uk> Reported-and-tested-by: Hans de Goede <jwrdegoede@fedoraproject.org> Reported-by: Barton Xu <tank.xuhan@gmail.com> Tested-by: Steffen Weber <steffen.weber@gmail.com> Tested-by: Arthur Chen <axchen@nvidia.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-15 08:41:35 +08:00
spin_lock_irqsave(&ec->lock, flags);
advance_transaction(ec);
spin_unlock_irqrestore(&ec->lock, flags);
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
return ACPI_INTERRUPT_HANDLED;
}
/* --------------------------------------------------------------------------
* Address Space Management
* -------------------------------------------------------------------------- */
static acpi_status
acpi_ec_space_handler(u32 function, acpi_physical_address address,
u32 bits, u64 *value64,
void *handler_context, void *region_context)
{
struct acpi_ec *ec = handler_context;
int result = 0, i, bytes = bits / 8;
u8 *value = (u8 *)value64;
if ((address > 0xFF) || !value || !handler_context)
return AE_BAD_PARAMETER;
if (function != ACPI_READ && function != ACPI_WRITE)
return AE_BAD_PARAMETER;
if (EC_FLAGS_MSI || bits > 8)
acpi_ec_burst_enable(ec);
for (i = 0; i < bytes; ++i, ++address, ++value)
result = (function == ACPI_READ) ?
acpi_ec_read(ec, address, value) :
acpi_ec_write(ec, address, *value);
if (EC_FLAGS_MSI || bits > 8)
acpi_ec_burst_disable(ec);
switch (result) {
case -EINVAL:
return AE_BAD_PARAMETER;
case -ENODEV:
return AE_NOT_FOUND;
case -ETIME:
return AE_TIME;
default:
return AE_OK;
}
}
/* --------------------------------------------------------------------------
* Driver Interface
* -------------------------------------------------------------------------- */
static acpi_status
ec_parse_io_ports(struct acpi_resource *resource, void *context);
static struct acpi_ec *make_acpi_ec(void)
{
struct acpi_ec *ec = kzalloc(sizeof(struct acpi_ec), GFP_KERNEL);
if (!ec)
return NULL;
ec->flags = 1 << EC_FLAGS_QUERY_PENDING;
mutex_init(&ec->mutex);
init_waitqueue_head(&ec->wait);
INIT_LIST_HEAD(&ec->list);
spin_lock_init(&ec->lock);
ACPI / EC: Fix issues related to the SCI_EVT handling This patch fixes 2 issues related to the draining behavior. But it doesn't implement the draining support, it only cleans up code so that further draining support is possible. The draining behavior is expected by some platforms (for example, Samsung) where SCI_EVT is set only once for a set of events and might be cleared for the very first QR_EC command issued after SCI_EVT is set. EC firmware on such platforms will return 0x00 to indicate "no outstanding event". Thus after seeing an SCI_EVT indication, EC driver need to fetch events until 0x00 returned (see acpi_ec_clear()). Issue 1 - acpi_ec_submit_query(): It's reported on Samsung laptops that SCI_EVT isn't checked when the transactions are advanced in ec_poll(), which leads to SCI_EVT triggering source lost: If no EC GPE IRQs are arrived after that, EC driver cannot detect this event and handle it. See comment 244/247 for kernel bugzilla 44161. This patch fixes this issue by moving SCI_EVT checks into advance_transaction(). So that SCI_EVT is checked each time we are going to handle the EC firmware indications. And this check will happen for both IRQ context and task context. Since after doing that, SCI_EVT is also checked after completing a transaction, ec_check_sci() and ec_check_sci_sync() can be removed. Issue 2 - acpi_ec_complete_query(): We expect to clear EC_FLAGS_QUERY_PENDING to allow queuing another draining QR_EC after writing a QR_EC command and before reading the event. After reading the event, SCI_EVT might be cleared by the firmware, thus it may not be possible to queue such a draining QR_EC at that time. But putting the EC_FLAGS_QUERY_PENDING clearing code after start_transaction() is wrong as there are chances that after start_transaction(), QR_EC can fail to be sent. If this happens, EC_FLAG_QUERY_PENDING will be cleared earlier. As a consequence, the draining QR_EC will also be queued earlier than expected. This patch also moves this code into advance_transaction() where QR_EC is just sent (ACPI_EC_COMMAND_POLL flagged) to fix this issue. Notes: 1. After introducing the 2 SCI_EVT related handlings into advance_transaction(), a next QR_EC can be queued right after writing the current QR_EC command and before reading the event. But this still hasn't implemented the draining behavior as the draining support requires: If a previous returned event value isn't 0x00, a draining QR_EC need to be issued even when SCI_EVT isn't set. 2. In this patch, acpi_os_execute() is also converted into a seperate work item to avoid invoking kmalloc() in the atomic context. We can do this because of the previous global lock fix. 3. Originally, EC_FLAGS_EVENT_PENDING is also used to avoid queuing up multiple work items (created by acpi_os_execute()), this can be covered by only using a single work item. But this patch still keeps this flag as there are different usages in the driver initialization steps relying on this flag. Link: https://bugzilla.kernel.org/show_bug.cgi?id=44161 Reported-by: Kieran Clancy <clancy.kieran@gmail.com> Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-01-14 19:28:47 +08:00
INIT_WORK(&ec->work, acpi_ec_gpe_poller);
return ec;
}
static acpi_status
acpi_ec_register_query_methods(acpi_handle handle, u32 level,
void *context, void **return_value)
{
char node_name[5];
struct acpi_buffer buffer = { sizeof(node_name), node_name };
struct acpi_ec *ec = context;
int value = 0;
acpi_status status;
status = acpi_get_name(handle, ACPI_SINGLE_NAME, &buffer);
if (ACPI_SUCCESS(status) && sscanf(node_name, "_Q%x", &value) == 1)
acpi_ec_add_query_handler(ec, value, handle, NULL, NULL);
return AE_OK;
}
static acpi_status
ec_parse_device(acpi_handle handle, u32 Level, void *context, void **retval)
{
acpi_status status;
unsigned long long tmp = 0;
struct acpi_ec *ec = context;
/* clear addr values, ec_parse_io_ports depend on it */
ec->command_addr = ec->data_addr = 0;
status = acpi_walk_resources(handle, METHOD_NAME__CRS,
ec_parse_io_ports, ec);
if (ACPI_FAILURE(status))
return status;
/* Get GPE bit assignment (EC events). */
/* TODO: Add support for _GPE returning a package */
status = acpi_evaluate_integer(handle, "_GPE", NULL, &tmp);
if (ACPI_FAILURE(status))
return status;
ec->gpe = tmp;
/* Use the global lock for all EC transactions? */
tmp = 0;
acpi_evaluate_integer(handle, "_GLK", NULL, &tmp);
ec->global_lock = tmp;
ec->handle = handle;
return AE_CTRL_TERMINATE;
}
static int ec_install_handlers(struct acpi_ec *ec)
{
acpi_status status;
if (test_bit(EC_FLAGS_HANDLERS_INSTALLED, &ec->flags))
return 0;
ACPI / EC: Fix several GPE handling issues by deploying ACPI_GPE_DISPATCH_RAW_HANDLER mode. This patch switches EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode where the GPE lock is not held for acpi_ec_gpe_handler() and the ACPICA internal GPE enabling/disabling/clearing operations are bypassed so that further improvements are possible with the GPE APIs. There are 2 strong reasons for deploying raw GPE handler mode in the EC driver: 1. Some hardware logics can control their interrupts via their own registers, so their interrupts can be disabled/enabled/acknowledged without using the super IRQ controller provided functions. While there is no mean (EC commands) for the EC driver to achieve this. 2. During suspending, the EC driver is still working for a while to complete the platform firmware provided functionailities using ec_poll() after all GPEs are disabled (see acpi_ec_block_transactions()), which means the EC driver will drive the EC GPE out of the GPE core's control. Without deploying the raw GPE handler mode, we can see many races between the EC driver and the GPE core due to the above restrictions: 1. There is a race condition due to ACPICA internal GPE disabling/clearing/enabling logics in acpi_ev_gpe_dispatch(): Orignally EC GPE is disabled (EN=0), cleared (STS=0) before invoking a GPE handler and re-enabled (EN=1) after invoking a GPE handler in acpi_ev_gpe_dispatch(). When re-enabling appears, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() ec_poll() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** EN=1 This race condition is the root cause of different issues on different silicon variations. A. Silicon variation A: On some platforms, GPE will be triggered due to "writing 1 to EN when STS=1". This is because both EN and STS lines are wired to the GPE trigger line. 1. Issue 1: We can see no-op acpi_ec_gpe_handler() invoked on such platforms. This is because: a. event pending B: An event can arrive after ACPICA's GPE clearing performed in acpi_ev_gpe_dispatch(), this event may fail to be detected by EC_SC read that is performed before its arrival; b. event handling B: The event can be handled in ec_poll() because EC lock is released after acpi_ec_gpe_handler() invocation; c. There is no code in ec_poll() to clear STS but the GPE can still be triggered by the EN=1 write performed in acpi_ev_finish_gpe(), this leads to a no-op EC GPE handler invocation; d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 1: If we removed GPE disabling/enabling code from acpi_ev_gpe_dispatch(), we could still see no-op GPE handlers triggered by the event arriving after the GPE clearing and before the GPE handling on both silicon variation A and B. This can only occur if the CPU is very slow (timing slice between STS=0 write and EC_SC read should be short enough before hardware sets another GPE indication). Thus this is very rare and is not what we need to fix. B. Silicon variation B: On other platforms, GPE may not be triggered due to "writing 1 to EN when STS=1". This is because only STS line is wired to the GPE trigger line. 2. Issue 2: We can see GPE loss on such platforms. This is because: a. event pending B vs. event handling A: An event can arrive after ACPICA's GPE handling performed in acpi_ev_gpe_dispatch(), or event pending C vs. event handling B: An event can arrive after Linux's GPE handling performed in ec_poll(), these events may fail to be detected by EC_SC read that is performed before their arrival; b. The GPE cannot be triggered by EN=1 write performed in acpi_ev_finish_gpe(); c. If no polling mechanism is implemented in the driver for the pending event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 2: On most platforms, there might be another rule that GPE may not be triggered due to "writing 1 to STS when STS=1 and EN=1". Then on silicon variation B, an even worse case is if the issue 2 event loss happens, further events may never trigger GPE again on such platforms due to being blocked by the current STS=1. Unless someone clears STS, all events have to be polled. 2. There is a race condition due to lacking in GPE status checking in EC driver: Originally, GPE status is checked in ACPICA core but not checked in the GPE handler. Thus since the status checking and handling is not locked, it can be interrupted by another handling path. ================================================================= (event pending A) ================================================================= acpi_ev_gpe_detect() ec_poll() if (EN==1 && STS==1) ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** acpi_ev_gpe_dispatch() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling B) Lock(EC) advance_transaction() EC_SC read Unlock(EC) ***************************************************************** 3. Issue 3: We can see no-op acpi_ec_gpe_handler() invoked on both silicon variation A and B. This is because: a. event pending A: An event can arrive to trigger an EC GPE and ACPICA checks it and is about to invoke the EC GPE handler; b. event handling A: The event can be handled in ec_poll() because EC lock is not held after the GPE status checking; c. event handling B: Then when the EC GPE handler is invoked, it becomes a no-op GPE handler invocation. d. As no-op GPE handler invocations are counted by the EC driver to trigger the command storming conditions, the wrong no-op GPE handler invocations thus can easily trigger wrong command storming conditions. Note 3: This no-op GPE handler invocation is rare because the time between the IRQ arrival and the acpi_ec_gpe_handler() invocation is less than the timeout value waited in ec_poll(). So most of the no-op GPE handler invocations are caused by the reason described in issue 1. 3. There is a race condition due to ACPICA internal GPE clearing logic in acpi_enable_gpe(): During runtime, acpi_enable_gpe() can be invoked by the EC storming prevention code. When it is invoked, GPE may be flagged (STS=1). ================================================================= (event pending A) ================================================================= acpi_ev_gpe_dispatch() acpi_ec_transaction() EN=0 STS=0 acpi_ec_gpe_handler() ***************************************************************** (event handling A) Lock(EC) advance_transaction() EC_SC read EC_SC handled Unlock(EC) ***************************************************************** EN=1 ? Lock(EC) Unlock(EC) ================================================================= (event pending B) ================================================================= acpi_enable_gpe() STS=0 EN=1 4. Issue 4: We can see GPE loss on both silicon variation A and B platforms. This is because: a. event pending B: An event can arrive right before ACPICA's GPE clearing performed in acpi_enable_gpe(); b. If the GPE is cleared when GPE is disabled, then EN=1 write in acpi_enable_gpe() cannot trigger this GPE; c. If no polling mechanism is implemented in the driver for this event (for example, SCI_EVT), this event is lost due to no GPE being triggered. Note 4: Currently we don't have this issue, but after we switch the EC driver into ACPI_GPE_DISPATCH_RAW_HANDLER mode, we need to take care of handling this because the EN=1 write in acpi_ev_gpe_dispatch() will be abandoned. There might be more race issues for the current GPE handler usages. This is because the EC IRQ's enabling/disabling/checking/clearing/handling operations should be locked by a single lock that is under the EC driver's control to achieve the serialization. Which means we need to invoke GPE APIs with EC driver's lock held and all ACPICA internal GPE operations related to the GPE handler should be abandoned. Invoking GPE APIs inside of the EC driver lock and bypassing ACPICA internal GPE operations requires the ACPI_GPE_DISPATCH_RAW_HANDLER mode where the same lock used by the APIs are released prior than invoking the handlers. Otherwise, we can see dead locks due to circular locking dependencies (see Reference below). This patch then switches the EC driver into the ACPI_GPE_DISPATCH_RAW_HANDLER mode so that it can perform correct GPE operations using the GPE APIs: 1. Bypasses EN modifications performed in acpi_ev_gpe_dispatch() by using acpi_install_gpe_raw_handler() and invoking all GPE APIs with EC spin lock held. This can fix issue 1 as it makes a non frequent GPE enabling/disabling environment. 2. Bypasses STS clearing performed in acpi_enable_gpe() by replacing acpi_enable_gpe()/acpi_disable_gpe() with acpi_set_gpe(). This can fix issue 4. And this can also help to fix issue 1 as it makes a no sudden GPE clearing environment when GPE is frequently enabled/disabled. 3. Ensures STS acknowledged before handling by invoking acpi_clear_gpe() in advance_transaction(). This can finally fix issue 1 even in a frequent GPE enabling/disabling environment. And this can also finally fix issue 3 when issue 2 is fixed. Note 3: GPE clearing is edge triggered W1C, which means we can clear it right before handling it. Since all EC GPE indications are handled in advance_transaction() by previous commits, we can now move GPE clearing into it to implement the correct GPE clearing. Note 4: We can use acpi_set_gpe() which is not shared GPE safer instead of acpi_enable_gpe()/acpi_disable_gpe() because EC GPE is not shared by other hardware, which is mentioned in the ACPI specification 5.0, 12.6 Interrupt Model: "OSPM driver treats this as an edge event (the EC SCI cannot be shared)". So we can stop using shared GPE safer APIs acpi_enable_gpe()/acpi_disable_gpe() in the EC driver. Otherwise cleanups need to be made in acpi_ev_enable_gpe() to bypass the GPE clearing logic before keeping acpi_enable_gpe(). This patch also invokes advance_transaction() when GPE is re-enabled in the task context which: 1. Ensures EN=1 can trigger GPE by checking and handling EC status register right after EN=1 writes. This can fix issue 2. After applying this patch, without frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() ec_poll() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 1 (event pending B) can arrive as a next GPE due to the previous IRQ context STS=0 write. And if it is handled by ec_poll() (event handling B), as it is also acknowledged by ec_poll(), the event pending for issue 2 (event pending C) can properly arrive as a next GPE after the task context STS=0 write. So no GPE will be lost and never triggered due to GPE clearing performed in the wrong position. And since all GPE handling is performed after a locked GPE status checking, we can hardly see no-op GPE handler invocations due to issue 1 and 3. We may still see no-op GPE handler invocations due to "Note 1", but as it is inevitable, it needn't be fixed. After applying this patch, with frequent GPE enablings considered: ================================================================= (event pending A) ================================================================= acpi_ec_gpe_handler() acpi_ec_transaction() ***************************************************************** (event handling A) Lock(EC) advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending B) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** ***************************************************************** (event handling B) Lock(EC) EN=1 if STS==1 advance_transaction() if STS==1 STS=0 EC_SC read ================================================================= (event pending C) ================================================================= EC_SC handled Unlock(EC) ***************************************************************** The event pending for issue 2 can be manually handled by advance_transaction(). And after the STS=0 write performed in the manual triggered advance_transaction(), GPE can always arrive. So no GPE will be lost due to frequent GPE disabling/enabling performed in the driver like issue 4. Note 5: It's ideally when EN=1 write occurred, an IRQ thread should be woken up to handle the GPE when the GPE was raised. But this requires the IRQ thread to contain the poller code for all EC GPE indications, while currently some of the indications are handled in the user tasks. It then is very hard for the code to determine whether a user task should be invoked or the poller work item should be scheduled. So we have to invoke advance_transaction() directly now and it leaves us such a restriction for the GPE re-enabling: it must be performed in the task context to avoid starving the GPEs. As a conclusion: we can see the EC GPE is always handled in serial after deploying the raw GPE handler mode: Lock(EC) if (STS==1) STS=0 EC_SC read EC_SC handled Unlock(EC) The EC driver specific lock is responsible to make the EC GPE handling processes serialized so that EC can handle its GPE from both IRQ and task contexts and the next IRQ can be ensured to arrive after this process. Note 6: We have many EC_FLAGS_MSI qurik users in the current driver. They all seem to be suffering from unexpected GPE triggering source lost. And they are false root caused to a timing issue. Since EC communication protocol has already flow control defined, timing shouldn't be the root cause, while this fix might be fixing the root cause of the old bugs. Link: https://lkml.org/lkml/2014/11/4/974 Link: https://lkml.org/lkml/2014/11/18/316 Link: https://www.spinics.net/lists/linux-acpi/msg54340.html Signed-off-by: Lv Zheng <lv.zheng@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-02-05 16:27:22 +08:00
status = acpi_install_gpe_raw_handler(NULL, ec->gpe,
ACPI_GPE_EDGE_TRIGGERED,
&acpi_ec_gpe_handler, ec);
if (ACPI_FAILURE(status))
return -ENODEV;
acpi_ec_start(ec, false);
status = acpi_install_address_space_handler(ec->handle,
ACPI_ADR_SPACE_EC,
&acpi_ec_space_handler,
NULL, ec);
if (ACPI_FAILURE(status)) {
if (status == AE_NOT_FOUND) {
/*
* Maybe OS fails in evaluating the _REG object.
* The AE_NOT_FOUND error will be ignored and OS
* continue to initialize EC.
*/
pr_err("Fail in evaluating the _REG object"
" of EC device. Broken bios is suspected.\n");
} else {
acpi_ec_stop(ec, false);
acpi_remove_gpe_handler(NULL, ec->gpe,
&acpi_ec_gpe_handler);
return -ENODEV;
}
}
set_bit(EC_FLAGS_HANDLERS_INSTALLED, &ec->flags);
return 0;
}
static void ec_remove_handlers(struct acpi_ec *ec)
{
if (!test_bit(EC_FLAGS_HANDLERS_INSTALLED, &ec->flags))
return;
acpi_ec_stop(ec, false);
if (ACPI_FAILURE(acpi_remove_address_space_handler(ec->handle,
ACPI_ADR_SPACE_EC, &acpi_ec_space_handler)))
pr_err("failed to remove space handler\n");
if (ACPI_FAILURE(acpi_remove_gpe_handler(NULL, ec->gpe,
&acpi_ec_gpe_handler)))
pr_err("failed to remove gpe handler\n");
clear_bit(EC_FLAGS_HANDLERS_INSTALLED, &ec->flags);
}
static int acpi_ec_add(struct acpi_device *device)
{
struct acpi_ec *ec = NULL;
int ret;
strcpy(acpi_device_name(device), ACPI_EC_DEVICE_NAME);
strcpy(acpi_device_class(device), ACPI_EC_CLASS);
/* Check for boot EC */
if (boot_ec &&
(boot_ec->handle == device->handle ||
boot_ec->handle == ACPI_ROOT_OBJECT)) {
ec = boot_ec;
boot_ec = NULL;
} else {
ec = make_acpi_ec();
if (!ec)
return -ENOMEM;
}
if (ec_parse_device(device->handle, 0, ec, NULL) !=
AE_CTRL_TERMINATE) {
kfree(ec);
return -EINVAL;
}
/* Find and register all query methods */
acpi_walk_namespace(ACPI_TYPE_METHOD, ec->handle, 1,
acpi_ec_register_query_methods, NULL, ec, NULL);
if (!first_ec)
first_ec = ec;
device->driver_data = ec;
ret = !!request_region(ec->data_addr, 1, "EC data");
WARN(!ret, "Could not request EC data io port 0x%lx", ec->data_addr);
ret = !!request_region(ec->command_addr, 1, "EC cmd");
WARN(!ret, "Could not request EC cmd io port 0x%lx", ec->command_addr);
pr_info("GPE = 0x%lx, I/O: command/status = 0x%lx, data = 0x%lx\n",
ec->gpe, ec->command_addr, ec->data_addr);
ret = ec_install_handlers(ec);
/* EC is fully operational, allow queries */
clear_bit(EC_FLAGS_QUERY_PENDING, &ec->flags);
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
/* Clear stale _Q events if hardware might require that */
if (EC_FLAGS_CLEAR_ON_RESUME)
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
acpi_ec_clear(ec);
return ret;
}
static int acpi_ec_remove(struct acpi_device *device)
{
struct acpi_ec *ec;
struct acpi_ec_query_handler *handler, *tmp;
if (!device)
return -EINVAL;
ec = acpi_driver_data(device);
ec_remove_handlers(ec);
mutex_lock(&ec->mutex);
list_for_each_entry_safe(handler, tmp, &ec->list, node) {
list_del(&handler->node);
kfree(handler);
}
mutex_unlock(&ec->mutex);
release_region(ec->data_addr, 1);
release_region(ec->command_addr, 1);
device->driver_data = NULL;
if (ec == first_ec)
first_ec = NULL;
kfree(ec);
return 0;
}
static acpi_status
ec_parse_io_ports(struct acpi_resource *resource, void *context)
{
struct acpi_ec *ec = context;
if (resource->type != ACPI_RESOURCE_TYPE_IO)
return AE_OK;
/*
* The first address region returned is the data port, and
* the second address region returned is the status/command
* port.
*/
if (ec->data_addr == 0)
ec->data_addr = resource->data.io.minimum;
else if (ec->command_addr == 0)
ec->command_addr = resource->data.io.minimum;
else
return AE_CTRL_TERMINATE;
return AE_OK;
}
int __init acpi_boot_ec_enable(void)
{
if (!boot_ec || test_bit(EC_FLAGS_HANDLERS_INSTALLED, &boot_ec->flags))
return 0;
if (!ec_install_handlers(boot_ec)) {
first_ec = boot_ec;
return 0;
}
return -EFAULT;
}
static const struct acpi_device_id ec_device_ids[] = {
{"PNP0C09", 0},
{"", 0},
};
/* Some BIOS do not survive early DSDT scan, skip it */
static int ec_skip_dsdt_scan(const struct dmi_system_id *id)
{
EC_FLAGS_SKIP_DSDT_SCAN = 1;
return 0;
}
/* ASUStek often supplies us with broken ECDT, validate it */
static int ec_validate_ecdt(const struct dmi_system_id *id)
{
EC_FLAGS_VALIDATE_ECDT = 1;
return 0;
}
/* MSI EC needs special treatment, enable it */
static int ec_flag_msi(const struct dmi_system_id *id)
{
pr_debug("Detected MSI hardware, enabling workarounds.\n");
EC_FLAGS_MSI = 1;
EC_FLAGS_VALIDATE_ECDT = 1;
return 0;
}
/*
* Clevo M720 notebook actually works ok with IRQ mode, if we lifted
* the GPE storm threshold back to 20
*/
static int ec_enlarge_storm_threshold(const struct dmi_system_id *id)
{
pr_debug("Setting the EC GPE storm threshold to 20\n");
ec_storm_threshold = 20;
return 0;
}
/*
* Acer EC firmware refuses to respond QR_EC when SCI_EVT is not set, for
* which case, we complete the QR_EC without issuing it to the firmware.
* https://bugzilla.kernel.org/show_bug.cgi?id=86211
*/
static int ec_flag_query_handshake(const struct dmi_system_id *id)
{
pr_debug("Detected the EC firmware requiring QR_EC issued when SCI_EVT set\n");
EC_FLAGS_QUERY_HANDSHAKE = 1;
return 0;
}
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
/*
* On some hardware it is necessary to clear events accumulated by the EC during
* sleep. These ECs stop reporting GPEs until they are manually polled, if too
* many events are accumulated. (e.g. Samsung Series 5/9 notebooks)
*
* https://bugzilla.kernel.org/show_bug.cgi?id=44161
*
* Ideally, the EC should also be instructed NOT to accumulate events during
* sleep (which Windows seems to do somehow), but the interface to control this
* behaviour is not known at this time.
*
* Models known to be affected are Samsung 530Uxx/535Uxx/540Uxx/550Pxx/900Xxx,
* however it is very likely that other Samsung models are affected.
*
* On systems which don't accumulate _Q events during sleep, this extra check
* should be harmless.
*/
static int ec_clear_on_resume(const struct dmi_system_id *id)
{
pr_debug("Detected system needing EC poll on resume.\n");
EC_FLAGS_CLEAR_ON_RESUME = 1;
return 0;
}
static struct dmi_system_id ec_dmi_table[] __initdata = {
{
ec_skip_dsdt_scan, "Compal JFL92", {
DMI_MATCH(DMI_BIOS_VENDOR, "COMPAL"),
DMI_MATCH(DMI_BOARD_NAME, "JFL92") }, NULL},
{
ec_flag_msi, "MSI hardware", {
DMI_MATCH(DMI_BIOS_VENDOR, "Micro-Star")}, NULL},
{
ec_flag_msi, "MSI hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "Micro-Star")}, NULL},
{
ec_flag_msi, "MSI hardware", {
DMI_MATCH(DMI_CHASSIS_VENDOR, "MICRO-Star")}, NULL},
{
ec_flag_msi, "MSI hardware", {
DMI_MATCH(DMI_CHASSIS_VENDOR, "MICRO-STAR")}, NULL},
{
ec_flag_msi, "Quanta hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "Quanta"),
DMI_MATCH(DMI_PRODUCT_NAME, "TW8/SW8/DW8"),}, NULL},
{
ec_flag_msi, "Quanta hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "Quanta"),
DMI_MATCH(DMI_PRODUCT_NAME, "TW9/SW9"),}, NULL},
{
ec_flag_msi, "Clevo W350etq", {
DMI_MATCH(DMI_SYS_VENDOR, "CLEVO CO."),
DMI_MATCH(DMI_PRODUCT_NAME, "W35_37ET"),}, NULL},
{
ec_validate_ecdt, "ASUS hardware", {
DMI_MATCH(DMI_BIOS_VENDOR, "ASUS") }, NULL},
{
ec_validate_ecdt, "ASUS hardware", {
DMI_MATCH(DMI_BOARD_VENDOR, "ASUSTeK Computer Inc.") }, NULL},
{
ec_enlarge_storm_threshold, "CLEVO hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "CLEVO Co."),
DMI_MATCH(DMI_PRODUCT_NAME, "M720T/M730T"),}, NULL},
{
ec_skip_dsdt_scan, "HP Folio 13", {
DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
DMI_MATCH(DMI_PRODUCT_NAME, "HP Folio 13"),}, NULL},
{
ec_validate_ecdt, "ASUS hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "ASUSTek Computer Inc."),
DMI_MATCH(DMI_PRODUCT_NAME, "L4R"),}, NULL},
ACPI / EC: Clear stale EC events on Samsung systems A number of Samsung notebooks (530Uxx/535Uxx/540Uxx/550Pxx/900Xxx/etc) continue to log events during sleep (lid open/close, AC plug/unplug, battery level change), which accumulate in the EC until a buffer fills. After the buffer is full (tests suggest it holds 8 events), GPEs stop being triggered for new events. This state persists on wake or even on power cycle, and prevents new events from being registered until the EC is manually polled. This is the root cause of a number of bugs, including AC not being detected properly, lid close not triggering suspend, and low ambient light not triggering the keyboard backlight. The bug also seemed to be responsible for performance issues on at least one user's machine. Juan Manuel Cabo found the cause of bug and the workaround of polling the EC manually on wake. The loop which clears the stale events is based on an earlier patch by Lan Tianyu (see referenced attachment). This patch: - Adds a function acpi_ec_clear() which polls the EC for stale _Q events at most ACPI_EC_CLEAR_MAX (currently 100) times. A warning is logged if this limit is reached. - Adds a flag EC_FLAGS_CLEAR_ON_RESUME which is set to 1 if the DMI system vendor is Samsung. This check could be replaced by several more specific DMI vendor/product pairs, but it's likely that the bug affects more Samsung products than just the five series mentioned above. Further, it should not be harmful to run acpi_ec_clear() on systems without the bug; it will return immediately after finding no data waiting. - Runs acpi_ec_clear() on initialisation (boot), from acpi_ec_add() - Runs acpi_ec_clear() on wake, from acpi_ec_unblock_transactions() References: https://bugzilla.kernel.org/show_bug.cgi?id=44161 References: https://bugzilla.kernel.org/show_bug.cgi?id=45461 References: https://bugzilla.kernel.org/show_bug.cgi?id=57271 References: https://bugzilla.kernel.org/attachment.cgi?id=126801 Suggested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Signed-off-by: Kieran Clancy <clancy.kieran@gmail.com> Reviewed-by: Lan Tianyu <tianyu.lan@intel.com> Reviewed-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Kieran Clancy <clancy.kieran@gmail.com> Tested-by: Juan Manuel Cabo <juanmanuel.cabo@gmail.com> Tested-by: Dennis Jansen <dennis.jansen@web.de> Tested-by: Maurizio D'Addona <mauritiusdadd@gmail.com> Tested-by: San Zamoyski <san@plusnet.pl> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-02-28 22:12:28 +08:00
{
ec_clear_on_resume, "Samsung hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "SAMSUNG ELECTRONICS CO., LTD.")}, NULL},
{
ec_flag_query_handshake, "Acer hardware", {
DMI_MATCH(DMI_SYS_VENDOR, "Acer"), }, NULL},
{},
};
int __init acpi_ec_ecdt_probe(void)
{
acpi_status status;
struct acpi_ec *saved_ec = NULL;
struct acpi_table_ecdt *ecdt_ptr;
boot_ec = make_acpi_ec();
if (!boot_ec)
return -ENOMEM;
/*
* Generate a boot ec context
*/
dmi_check_system(ec_dmi_table);
status = acpi_get_table(ACPI_SIG_ECDT, 1,
(struct acpi_table_header **)&ecdt_ptr);
if (ACPI_SUCCESS(status)) {
pr_info("EC description table is found, configuring boot EC\n");
boot_ec->command_addr = ecdt_ptr->control.address;
boot_ec->data_addr = ecdt_ptr->data.address;
boot_ec->gpe = ecdt_ptr->gpe;
boot_ec->handle = ACPI_ROOT_OBJECT;
acpi_get_handle(ACPI_ROOT_OBJECT, ecdt_ptr->id,
&boot_ec->handle);
/* Don't trust ECDT, which comes from ASUSTek */
if (!EC_FLAGS_VALIDATE_ECDT)
goto install;
saved_ec = kmemdup(boot_ec, sizeof(struct acpi_ec), GFP_KERNEL);
if (!saved_ec)
return -ENOMEM;
/* fall through */
}
if (EC_FLAGS_SKIP_DSDT_SCAN) {
kfree(saved_ec);
return -ENODEV;
}
/* This workaround is needed only on some broken machines,
* which require early EC, but fail to provide ECDT */
pr_debug("Look up EC in DSDT\n");
status = acpi_get_devices(ec_device_ids[0].id, ec_parse_device,
boot_ec, NULL);
/* Check that acpi_get_devices actually find something */
if (ACPI_FAILURE(status) || !boot_ec->handle)
goto error;
if (saved_ec) {
/* try to find good ECDT from ASUSTek */
if (saved_ec->command_addr != boot_ec->command_addr ||
saved_ec->data_addr != boot_ec->data_addr ||
saved_ec->gpe != boot_ec->gpe ||
saved_ec->handle != boot_ec->handle)
pr_info("ASUSTek keeps feeding us with broken "
"ECDT tables, which are very hard to workaround. "
"Trying to use DSDT EC info instead. Please send "
"output of acpidump to linux-acpi@vger.kernel.org\n");
kfree(saved_ec);
saved_ec = NULL;
} else {
/* We really need to limit this workaround, the only ASUS,
* which needs it, has fake EC._INI method, so use it as flag.
* Keep boot_ec struct as it will be needed soon.
*/
if (!dmi_name_in_vendors("ASUS") ||
!acpi_has_method(boot_ec->handle, "_INI"))
return -ENODEV;
}
install:
if (!ec_install_handlers(boot_ec)) {
first_ec = boot_ec;
return 0;
}
error:
kfree(boot_ec);
kfree(saved_ec);
boot_ec = NULL;
return -ENODEV;
}
static struct acpi_driver acpi_ec_driver = {
.name = "ec",
.class = ACPI_EC_CLASS,
.ids = ec_device_ids,
.ops = {
.add = acpi_ec_add,
.remove = acpi_ec_remove,
},
};
int __init acpi_ec_init(void)
{
int result = 0;
/* Now register the driver for the EC */
result = acpi_bus_register_driver(&acpi_ec_driver);
if (result < 0)
return -ENODEV;
return result;
}
/* EC driver currently not unloadable */
#if 0
static void __exit acpi_ec_exit(void)
{
acpi_bus_unregister_driver(&acpi_ec_driver);
}
#endif /* 0 */