A Root Port's AER structure (rpc) contains a queue of events. aer_irq()
enqueues AER status information and schedules aer_isr() to dequeue and
process it. When we remove a device, aer_remove() waits for the queue to
be empty, then frees the rpc struct.
But aer_isr() references the rpc struct after dequeueing and possibly
emptying the queue, which can cause a use-after-free error as in the
following scenario with two threads, aer_isr() on the left and a
concurrent aer_remove() on the right:
Thread A Thread B
-------- --------
aer_irq():
rpc->prod_idx++
aer_remove():
wait_event(rpc->prod_idx == rpc->cons_idx)
# now blocked until queue becomes empty
aer_isr(): # ...
rpc->cons_idx++ # unblocked because queue is now empty
... kfree(rpc)
mutex_unlock(&rpc->rpc_mutex)
To prevent this problem, use flush_work() to wait until the last scheduled
instance of aer_isr() has completed before freeing the rpc struct in
aer_remove().
I reproduced this use-after-free by flashing a device FPGA and
re-enumerating the bus to find the new device. With SLUB debug, this
crashes with 0x6b bytes (POISON_FREE, the use-after-free magic number) in
GPR25:
pcieport 0000:00:00.0: AER: Multiple Corrected error received: id=0000
Unable to handle kernel paging request for data at address 0x27ef9e3e
Workqueue: events aer_isr
GPR24: dd6aa000 6b6b6b6b 605f8378 605f8360 d99b12c0 604fc674 606b1704 d99b12c0
NIP [602f5328] pci_walk_bus+0xd4/0x104
[bhelgaas: changelog, stable tag]
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: stable@vger.kernel.org
One PCI bus reset function to rule them all.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The function aer_recover_queue() is a public interface and the
severity argument uses #defines that are in the private header
pci/pcie/aer/aerdrv.h.
This patch moves the #defines from pci/pcie/aer/aerdrv.h to
include/linux/aer.h.
[bhelgaas: split "remove 'extern' from declarations" to another patch]
Signed-off-by: Betty Dall <betty.dall@hp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
We had an inconsistent mix of using and omitting the "extern" keyword
on function declarations in header files. This removes them all.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
When an error is detected on a PCIe device which does not have an
AER-aware driver, prevent AER infrastructure from reporting
successful error recovery.
This is because the report_error_detected() function that gets
called in the first phase of recovery process allows forward
progress even when the driver for the device does not have AER
capabilities. It seems that all callbacks (in pci_error_handlers
structure) registered by drivers that gets called during error
recovery are not mandatory. So the intention of the infrastructure
design seems to be to allow forward progress even when a specific
callback has not been registered by a driver. However, if error
handler structure itself has not been registered, it doesn't make
sense to allow forward progress.
As a result of the current design, in the case of a single device
having an AER-unaware driver or in the case of any function in a
multi-function card having an AER-unaware driver, a successful
recovery is reported.
Typical scenario this happens is when a PCI device is detached
from a KVM host and the pci-stub driver on the host claims the
device. The pci-stub driver does not have error handling capabilities
but the AER infrastructure still reports that the device recovered
successfully.
The changes proposed here leaves the device(s)in an unrecovered state
if the driver for the device or for any device in the subtree
does not have error handler structure registered. This reflects
the true state of the device and prevents any partial recovery (or no
recovery at all) reported as successful.
[bhelgaas: changelog]
Signed-off-by: Vijay Mohan Pandarathil <vijaymohan.pandarathil@hp.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Linas Vepstas <linasvepstas@gmail.com>
Reviewed-by: Myron Stowe <myron.stowe@redhat.com>
In the commit 28eb5f2, aer_osc_setup is removed but corresponding
definiton information in the aerdrv.h is missed.
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Chen Gong <gong.chen@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The AER error information printing support is implemented in
drivers/pci/pcie/aer/aer_print.c. So some string constants, functions
and macros definitions can be re-used without being exported.
The original PCIe AER error information printing function is not
re-used directly because the overall format is quite different. And
changing the original printing format may make some original users'
scripts broken.
Signed-off-by: Huang Ying <ying.huang@intel.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
CC: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Move the evaluation of acpi_pci_osc_control_set() (to request control of
PCI Express native features) into acpi_pci_root_add() to avoid calling
it many times for the same root complex with the same arguments.
Additionally, check if all of the requisite _OSC support bits are set
before calling acpi_pci_osc_control_set() for a given root complex.
References: https://bugzilla.kernel.org/show_bug.cgi?id=20232
Reported-by: Ozan Caglayan <ozan@pardus.org.tr>
Tested-by: Ozan Caglayan <ozan@pardus.org.tr>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
There is a design issue related to PCIe AER and _OSC that the BIOS
may be asked to grant control of the AER service even if some
Hardware Error Source Table (HEST) entries contain information
meaning that the BIOS really should control it. Namely,
pcie_port_acpi_setup() calls pcie_aer_get_firmware_first() that
determines whether or not the AER service should be controlled by
the BIOS on the basis of the HEST information for the given PCIe
port. The BIOS is asked to grant control of the AER service for
a PCIe Root Complex if pcie_aer_get_firmware_first() returns 'false'
for at least one root port in that complex, even if all of the other
root ports' HEST entries have the FIRMWARE_FIRST flag set (and none
of them has the GLOBAL flag set). However, if the AER service is
controlled by the kernel, that may interfere with the BIOS' handling
of the error sources having the FIRMWARE_FIRST flag. Moreover,
there may be PCIe endpoints that have the FIRMWARE_FIRST flag set in
HEST and are attached to the root ports in question, in which case it
also may be unsafe to ask the BIOS for control of the AER service.
For this reason, introduce a function checking if there's at least
one PCIe-related HEST entry with the FIRMWARE_FIRST flag set and
disable the native AER service altogether if this function returns
'true'.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Now, a dedicated HEST tabling parsing code is used for PCIE AER
firmware_first setup. It is rebased on general HEST tabling parsing
code of APEI. The firmware_first setup code is moved from PCI core to
AER driver too, because it is only AER related.
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Len Brown <len.brown@intel.com>
I noticed that when I inject a fatal error to an endpoint via
aer-inject, aer_root_reset() is called as reset_link for a
downstream port at upstream of the endpoint:
pcieport 0000:00:06.0: AER: Uncorrected (Fatal) error received: id=5401
:
pcieport 0000:52:02.0: Root Port link has been reset
It externally appears to be working, but internally issues some
accesses to PCI_ERR_ROOT_COMMAND/STATUS registers that is for
root port so not available on downstream port.
This patch introduces default_downstream_reset_link that is
a version of aer_root_reset() with no accesses to root port's
register. It is used for downstream ports that has no reset_link
function its specific.
This patch also updates related description in pcieaer-howto.txt.
Some minor fixes are included.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Make it clear that we only interest in 2 *_RCV bits.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
These functions are only called from init/remove path of aerdrv,
so move them from aerdrv_core.c to aerdrv.c, to make them static.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This cleanup solves some minor naming issues by removing unuseful
function aer_delete_rootport() and by renaming disable_root_aer()
to aer_disable_rootport().
- Inconsistent location of alloc & free:
The struct rpc is allocated in aer_alloc_rpc() at aerdrv.c
while it is implicitly freed in aer_delete_rootport() at
aerdrv_core.c.
- Inconsistent function name:
It makes a bit confusion that aer_delete_rootport() is seemed
to be paired with aer_enable_rootport(), i.e. there is neither
"add" against "delete" nor "disable" against "enable".
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Multiple bits might be set in the Uncorrectable Error Status
register. But aer_print_error_source() only report a error of
the lowest bit set in the error status register.
So print strings for all bits unmasked and set.
And check First Error Pointer to mark the error occured first.
This FEP is not valid when the corresponing bit of the Uncorrectable
Error Status register is not set, or unimplemented or undefined.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
ERR_{,UN}CORRECTABLE_ERROR_MASK are set of error bits which linux know,
set of PCI_ERR_COR_* and PCI_ERR_UNC_* defined in linux/pci_regs.h.
This masks make aerdrv not to report errors of unknown bit, while aerdrv
have ability to report such undefined errors as "Unknown Error Bit %2d".
OTOH aerdrv_errprint does not have any check of setting in mask register.
So it could report masked wrong error by finding bit in status without
knowing that the bit is masked in the mask register.
This patch changes aerdrv to use mask state in mask register propely
instead of defined/hardcoded ERR_{,UN}CORRECTABLE_ERROR_MASK.
This change prevents aerdrv from reporting masked error, and also enable
reporting unknown errors.
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
When a root port receives the same errors more than once before the
kernel process them, the Multiple Error Messages Received flags are set
by hardware. Because the root port could only save one kind of
correctable error source id and another uncorrectable error source id at
the same time, the second message sender id is lost if the 2 messages
are sent from 2 different devices. This patch makes the kernel search
all devices under the root port when multiple messages are received.
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Zhang Yanmin <yanmin_zhang@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
When the bus id part of error source id is equal to 0 or nosourceid=1,
make the kernel probe the AER status registers of all devices under the
root port to find the initial error reporter.
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Zhang Yanmin <yanmin_zhang@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This is used by PCIE AER error injection to fake an PCI AER interrupt.
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
When an upstream port reports an AER error to root port, kernel
starts error recovery procedures. The default return value of
function pcie_portdrv_slot_reset is PCI_ERS_RESULT_NONE. If all
port service drivers of the downstream port under the upstream
port have no slot_reset method in pci_error_handlers, AER recovery
would stop without resume. Below patch against 2.6.30-rc3 fixes it.
Signed-off-by: Zhang Yanmin <yanmin.zhang@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Below patch fixes aer driver error information and enables aer driver
although CONFIG_ACPI=n.
As a matter of fact, the new patch is created from below 2 patches plus
a minor patch apply fuzz fixing. Because the second patch fixed a compilation
error introduced by the first patch, I merge them to facilitate bisect.
1) http://marc.info/?l=linux-kernel&m=117783233918191&w=2;
2) http://marc.info/?l=linux-mm-commits&m=118046936720790&w=2
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
First thing mm.h does is including sched.h solely for can_do_mlock() inline
function which has "current" dereference inside. By dealing with can_do_mlock()
mm.h can be detached from sched.h which is good. See below, why.
This patch
a) removes unconditional inclusion of sched.h from mm.h
b) makes can_do_mlock() normal function in mm/mlock.c
c) exports can_do_mlock() to not break compilation
d) adds sched.h inclusions back to files that were getting it indirectly.
e) adds less bloated headers to some files (asm/signal.h, jiffies.h) that were
getting them indirectly
Net result is:
a) mm.h users would get less code to open, read, preprocess, parse, ... if
they don't need sched.h
b) sched.h stops being dependency for significant number of files:
on x86_64 allmodconfig touching sched.h results in recompile of 4083 files,
after patch it's only 3744 (-8.3%).
Cross-compile tested on
all arm defconfigs, all mips defconfigs, all powerpc defconfigs,
alpha alpha-up
arm
i386 i386-up i386-defconfig i386-allnoconfig
ia64 ia64-up
m68k
mips
parisc parisc-up
powerpc powerpc-up
s390 s390-up
sparc sparc-up
sparc64 sparc64-up
um-x86_64
x86_64 x86_64-up x86_64-defconfig x86_64-allnoconfig
as well as my two usual configs.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pass the work_struct pointer to the work function rather than context data.
The work function can use container_of() to work out the data.
For the cases where the container of the work_struct may go away the moment the
pending bit is cleared, it is made possible to defer the release of the
structure by deferring the clearing of the pending bit.
To make this work, an extra flag is introduced into the management side of the
work_struct. This governs auto-release of the structure upon execution.
Ordinarily, the work queue executor would release the work_struct for further
scheduling or deallocation by clearing the pending bit prior to jumping to the
work function. This means that, unless the driver makes some guarantee itself
that the work_struct won't go away, the work function may not access anything
else in the work_struct or its container lest they be deallocated.. This is a
problem if the auxiliary data is taken away (as done by the last patch).
However, if the pending bit is *not* cleared before jumping to the work
function, then the work function *may* access the work_struct and its container
with no problems. But then the work function must itself release the
work_struct by calling work_release().
In most cases, automatic release is fine, so this is the default. Special
initiators exist for the non-auto-release case (ending in _NAR).
Signed-Off-By: David Howells <dhowells@redhat.com>
Patch 3 implements the core part of PCI-Express AER and aerdrv
port service driver.
When a root port service device is probed, the aerdrv will call
request_irq to register irq handler for AER error interrupt.
When a device sends an PCI-Express error message to the root port,
the root port will trigger an interrupt, by either MSI or IO-APIC,
then kernel would run the irq handler. The handler collects root
error status register and schedules a work. The work will call
the core part to process the error based on its type
(Correctable/non-fatal/fatal).
As for Correctable errors, the patch chooses to just clear the correctable
error status register of the device.
As for the non-fatal error, the patch follows generic PCI error handler
rules to call the error callback functions of the endpoint's driver. If
the device is a bridge, the patch chooses to broadcast the error to
downstream devices.
As for the fatal error, the patch resets the pci-express link and
follows generic PCI error handler rules to call the error callback
functions of the endpoint's driver. If the device is a bridge, the patch
chooses to broadcast the error to downstream devices.
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Sebastian Andrzej Siewior