Currently, FADump is only supported on pSeries but that is going to
change soon with FADump support being added on PowerNV platform. So,
move rtas specific definitions to platform code to allow FADump
to have multiple platforms support.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/156821328494.5656.16219929140866195511.stgit@hbathini.in.ibm.com
The figures depicting FADump's (Firmware-Assisted Dump) memory layout
are missing some finer details like different memory regions and what
they represent. Improve the documentation by updating those details.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/156821322070.5656.8194734198500730487.stgit@hbathini.in.ibm.com
Add helper functions to setup & free CPU notes buffer and to find if a
given memory area is contiguous. Also, use boolean as return type for
the function that finds if boot memory area is contiguous. While at
it, save the virtual address of CPU notes buffer instead of physical
address as virtual address is used often.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/156821318971.5656.9281936950510635858.stgit@hbathini.in.ibm.com
Though asm/fadump.h is meant to be used by other components dealing
with FADump, it also has macros/definitions internal to FADump code.
Move them to a new header file used within FADump code. This also
makes way for refactoring platform specific FADump code.
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/156821313134.5656.6597770626574392140.stgit@hbathini.in.ibm.com
This slightly improves the prom_init_check rule.
[1] Avoid needless check
Currently, prom_init_check.sh is invoked every time you run 'make'
even if you have changed nothing in prom_init.c. With this commit,
the script is re-run only when prom_init.o is recompiled.
[2] Beautify the build log
Currently, the O= build shows the absolute path to the script:
CALL /abs/path/to/source/of/linux/arch/powerpc/kernel/prom_init_check.sh
With this commit, it is always a relative path to the timestamp file:
PROMCHK arch/powerpc/kernel/prom_init_check
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190912074037.13813-1-yamada.masahiro@socionext.com
Some of the templates used for KVM patching are only used on certain
platforms, but currently they are always built-in, fix that.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911115746.12433-4-mpe@ellerman.id.au
All the code in kvm.c can be marked __init. Most of it is already
inlined into the initcall, but not all. So instead of relying on the
inlining, mark it all as __init. This saves ~280 bytes of text for my
configuration.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911115746.12433-3-mpe@ellerman.id.au
In some configurations of KVM, guests binary patch themselves to
avoid/reduce trapping into the hypervisor. For some instructions this
requires replacing one instruction with a sequence of instructions.
For those cases we need to write the sequence of instructions
somewhere and then patch the location of the original instruction to
branch to the sequence. That requires that the location of the
sequence be within 32MB of the original instruction.
The current solution for this is that we create a 1MB array in BSS,
write sequences into there, and then free the remainder of the array.
This has a few problems:
- it confuses kmemleak.
- it confuses lockdep.
- it requires mapping kvm_tmp executable, which can cause adjacent
areas to also be mapped executable if we're using 16M pages for the
linear mapping.
- the 32MB limit can be exceeded if the kernel is big enough,
especially with STRICT_KERNEL_RWX enabled, which then prevents the
patching from working at all.
We can fix all those problems by making kvm_tmp just a region of
regular .text. However currently it's 1MB in size, and we don't want
to waste 1MB of text. In practice however I only see ~30KB of kvm_tmp
being used even for an allyes_config. So shrink kvm_tmp to 64K, which
ought to be enough for everyone, and move it into .text.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911115746.12433-1-mpe@ellerman.id.au
The builds breaks when IOMMU_API=n, eg. skiroot_defconfig:
arch/powerpc/platforms/powernv/npu-dma.c:96:28: error: 'get_gpu_pci_dev_and_pe' defined but not used
arch/powerpc/platforms/powernv/npu-dma.c:126:13: error: 'pnv_npu_set_window' defined but not used
Fixes: b4d37a7b69 ("powerpc/powernv: Remove unused pnv_npu_try_dma_set_bypass() function")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The build breaks when STACKTRACE=n, eg. skiroot_defconfig:
arch/powerpc/kernel/eeh_event.c:124:23: error: implicit declaration of function 'stack_trace_save'
Fix it with some ifdefs for now.
Fixes: 25baf3d816 ("powerpc/eeh: Defer printing stack trace")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There's a bug in skiboot that causes the OPAL_XIVE_ALLOCATE_IRQ call
to return the 32-bit value 0xffffffff when OPAL has run out of IRQs.
Unfortunatelty, OPAL return values are signed 64-bit entities and
errors are supposed to be negative. If that happens, the linux code
confusingly treats 0xffffffff as a valid IRQ number and panics at some
point.
A fix was recently merged in skiboot:
e97391ae2bb5 ("xive: fix return value of opal_xive_allocate_irq()")
but we need a workaround anyway to support older skiboots already
in the field.
Internally convert 0xffffffff to OPAL_RESOURCE which is the usual error
returned upon resource exhaustion.
Cc: stable@vger.kernel.org # v4.12+
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/156821713818.1985334.14123187368108582810.stgit@bahia.lan
prep_irq_for_idle() is intended to be called before entering
H_CEDE (and it is used by the pseries cpuidle driver). However the
default pseries idle routine does not call it, leading to mismanaged
lazy irq state when the cpuidle driver isn't in use. Manifestations of
this include:
* Dropped IPIs in the time immediately after a cpu comes
online (before it has installed the cpuidle handler), making the
online operation block indefinitely waiting for the new cpu to
respond.
* Hitting this WARN_ON in arch_local_irq_restore():
/*
* We should already be hard disabled here. We had bugs
* where that wasn't the case so let's dbl check it and
* warn if we are wrong. Only do that when IRQ tracing
* is enabled as mfmsr() can be costly.
*/
if (WARN_ON_ONCE(mfmsr() & MSR_EE))
__hard_irq_disable();
Call prep_irq_for_idle() from pseries_lpar_idle() and honor its
result.
Fixes: 363edbe261 ("powerpc: Default arch idle could cede processor on pseries")
Signed-off-by: Nathan Lynch <nathanl@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190910225244.25056-1-nathanl@linux.ibm.com
If watchpoint exception is generated by larx/stcx instructions, the
reservation created by larx gets lost while handling exception, and
thus stcx instruction always fails. Generally these instructions are
used in a while(1) loop, for example spinlocks. And because stcx
never succeeds, it loops forever and ultimately hangs the system.
Note that ptrace anyway works in one-shot mode and thus for ptrace
we don't change the behaviour. It's up to ptrace user to take care
of this.
Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190910131513.30499-1-ravi.bangoria@linux.ibm.com
We have OPAL_MSG_PRD message type to pass prd related messages from
OPAL to `opal-prd`. It can handle messages upto 64 bytes. We have a
requirement to send bigger than 64 bytes of data from OPAL to
`opal-prd`. Lets add new message type (OPAL_MSG_PRD2) to pass bigger
data.
Signed-off-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com>
[mpe: Make the error string clear that it's the PRD2 event that failed]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190826065701.8853-2-hegdevasant@linux.vnet.ibm.com
Use "opal-msg-size" device tree property to allocate memory for
"opal_msg".
Signed-off-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com>
[mpe: s/uint32_t/u32/ and mark opal_msg_size as __ro_after_init]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190826065701.8853-1-hegdevasant@linux.vnet.ibm.com
Neither pnv_npu_try_dma_set_bypass() nor the pnv_npu_dma_set_32() and
pnv_npu_dma_set_bypass() helpers called by it are used anywhere in the
kernel tree, so remove them.
mpe: They're unused since 2d6ad41b2c ("powerpc/powernv: use the
generic iommu bypass code") removed the last usage.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903165147.11099-1-hch@lst.de
For sizes lesser than 128 bytes, the code branches out early without saving
the stack frame, which when restored later drops frame of the caller.
Tested-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Santosh Sivaraj <santosh@fossix.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903214359.23887-1-santosh@fossix.org
Commit 2874c5fd28 ("treewide: Replace GPLv2 boilerplate/reference with
SPDX - rule 152") left an empty comment in machdep.h, as the boilerplate
was the only text in the comment. Remove the empty comment.
Signed-off-by: Jordan Niethe <jniethe5@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190813051212.6387-1-jniethe5@gmail.com
Commit <684d984038aa> ('powerpc/powernv: Add debugfs interface for
imc-mode and imc') added debugfs interface for the nest imc pmu
devices to support changing of different ucode modes. Primarily adding
this capability for debug. But when doing so, the code did not
consider the case of cpu-less nodes. So when reading the _cmd_ or
_mode_ file of a cpu-less node will create this crash.
Faulting instruction address: 0xc0000000000d0d58
Oops: Kernel access of bad area, sig: 11 [#1]
...
CPU: 67 PID: 5301 Comm: cat Not tainted 5.2.0-rc6-next-20190627+ #19
NIP: c0000000000d0d58 LR: c00000000049aa18 CTR:c0000000000d0d50
REGS: c00020194548f9e0 TRAP: 0300 Not tainted (5.2.0-rc6-next-20190627+)
MSR: 9000000000009033 <SF,HV,EE,ME,IR,DR,RI,LE> CR:28022822 XER: 00000000
CFAR: c00000000049aa14 DAR: 000000000003fc08 DSISR:40000000 IRQMASK: 0
...
NIP imc_mem_get+0x8/0x20
LR simple_attr_read+0x118/0x170
Call Trace:
simple_attr_read+0x70/0x170 (unreliable)
debugfs_attr_read+0x6c/0xb0
__vfs_read+0x3c/0x70
vfs_read+0xbc/0x1a0
ksys_read+0x7c/0x140
system_call+0x5c/0x70
Patch fixes the issue with a more robust check for vbase to NULL.
Before patch, ls output for the debugfs imc directory
# ls /sys/kernel/debug/powerpc/imc/
imc_cmd_0 imc_cmd_251 imc_cmd_253 imc_cmd_255 imc_mode_0 imc_mode_251 imc_mode_253 imc_mode_255
imc_cmd_250 imc_cmd_252 imc_cmd_254 imc_cmd_8 imc_mode_250 imc_mode_252 imc_mode_254 imc_mode_8
After patch, ls output for the debugfs imc directory
# ls /sys/kernel/debug/powerpc/imc/
imc_cmd_0 imc_cmd_8 imc_mode_0 imc_mode_8
Actual bug here is that, we have two loops with potentially different
loop counts. That is, in imc_get_mem_addr_nest(), loop count is
obtained from the dt entries. But in case of export_imc_mode_and_cmd(),
loop was based on for_each_nid() count. Patch fixes the loop count in
latter based on the struct mem_info. Ideally it would be better to
have array size in struct imc_pmu.
Fixes: 684d984038 ('powerpc/powernv: Add debugfs interface for imc-mode and imc')
Reported-by: Qian Cai <cai@lca.pw>
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190827101635.6942-1-maddy@linux.vnet.ibm.com
Introduce two options to control the use of the tlbie instruction. A
boot time option which completely disables the kernel using the
instruction, this is currently incompatible with HASH MMU, KVM, and
coherent accelerators.
And a debugfs option can be switched at runtime and avoids using tlbie
for invalidating CPU TLBs for normal process and kernel address
mappings. Coherent accelerators are still managed with tlbie, as will
KVM partition scope translations.
Cross-CPU TLB flushing is implemented with IPIs and tlbiel. This is a
basic implementation which does not attempt to make any optimisation
beyond the tlbie implementation.
This is useful for performance testing among other things. For example
in certain situations on large systems, using IPIs may be faster than
tlbie as they can be directed rather than broadcast. Later we may also
take advantage of the IPIs to do more interesting things such as trim
the mm cpumask more aggressively.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190902152931.17840-7-npiggin@gmail.com
The various translation structure invalidations performed in early boot
when the MMU is off are not required, because everything is invalidated
immediately before a CPU first enables its MMU (see early_init_mmu
and early_init_mmu_secondary).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190902152931.17840-6-npiggin@gmail.com
Radix guests are responsible for managing their own translation caches,
so make them match bare metal radix and hash, and make each CPU flush
all its translations right before enabling its MMU.
Radix guests may not flush partition scope translations, so in
tlbiel_all, make these flushes conditional on CPU_FTR_HVMODE. Process
scope translations are the only type visible to the guest.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190902152931.17840-5-npiggin@gmail.com
There should be no functional changes.
- Use calls to existing radix_tlb.c functions in flush_partition.
- Rename radix__flush_tlb_lpid to radix__flush_all_lpid and similar,
because they flush everything, matching flush_all_mm rather than
flush_tlb_mm for the lpid.
- Remove some unused radix_tlb.c flush primitives.
Signed-off: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190902152931.17840-3-npiggin@gmail.com
This callback is only required because the partition table init comes
before process table allocation on powernv (aka bare metal aka native).
Change the order to allocate the process table first, and remove the
callback.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190902152931.17840-2-npiggin@gmail.com
Use the new eeh_dev_check and eeh_dev_break interfaces to test EEH
recovery. Historically this has been done manually using platform specific
EEH error injection facilities (e.g. via RTAS). However, documentation on
how to use these facilities is haphazard at best and non-existent at worst
so it's hard to develop a cross-platform test.
The new debugfs interfaces allow the kernel to handle the platform specific
details so we can write a more generic set of sets. This patch adds the
most basic of recovery tests where:
a) Errors are injected and recovered from sequentially,
b) Errors are not injected into PCI-PCI bridges, such as PCIe switches.
c) Errors are only injected into device function zero.
d) No errors are injected into Virtual Functions.
a), b) and c) are largely due to limitations of Linux's EEH support. EEH
recovery is serialised in the EEH recovery thread which forces a).
Similarly, multi-function PCI devices are almost always grouped into the
same PE so injecting an error on one function exercises the same code
paths. c) is because we currently more or less ignore PCI bridges during
recovery and assume that the recovered topology will be the same as the
original.
d) is due to the limits of the eeh_dev_break interface. With the current
implementation we can't inject an error into a specific VF without
potentially causing additional errors on other VFs. Due to the serialised
recovery process we might end up timing out waiting for another function to
recover before the function of interest is recovered. The platform specific
error injection facilities are finer-grained and allow this capability, but
doing that requires working out how to use those facilities first.
Basicly, it's better than nothing and it's a base to build on.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-15-oohall@gmail.com
Add an interface to debugfs for generating an EEH event on a given device.
This works by disabling memory accesses to and from the device by setting
the PCI_COMMAND register (or the VF Memory Space Enable on the parent PF).
This is a somewhat portable alternative to using the platform specific
error injection mechanisms since those tend to be either hard to use, or
straight up broken. For pseries the interfaces also requires the use of
/dev/mem which is probably going to go away in a post-LOCKDOWN world
(and it's a horrific hack to begin with) so moving to a kernel-provided
interface makes sense and provides a sane, cross-platform interface for
userspace so we can write more generic testing scripts.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-14-oohall@gmail.com
Detecting an frozen EEH PE usually occurs when an MMIO load returns a 0xFFs
response. When performing EEH testing using the EEH error injection feature
available on some platforms there is no simple way to kick-off the kernel's
recovery process since any accesses from userspace (usually /dev/mem) will
bypass the MMIO helpers in the kernel which check if a 0xFF response is due
to an EEH freeze or not.
If a device contains a 0xFF byte in it's config space it's possible to
trigger the recovery process via config space read from userspace, but this
is not a reliable method. If a driver is bound to the device an in use it
will frequently trigger the MMIO check, but this is also inconsistent.
To solve these problems this patch adds a debugfs file called
"eeh_dev_check" which accepts a <domain>:<bus>:<dev>.<fn> string and runs
eeh_dev_check_failure() on it. This is the same check that's done when the
kernel gets a 0xFF result from an config or MMIO read with the added
benifit that it can be reliably triggered from userspace.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-13-oohall@gmail.com
I am the RAS team. Hear me roar.
Roar.
On a more serious note, being able to locate failed devices can be helpful.
Set the attention indicator if the slot supports it once we've determined
the device is present and only clear it if the device is fully recovered.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-12-oohall@gmail.com
pnv_php is generally used with PCIe bridges which provide a native
interface for setting the attention and power indicator LEDs. Wire up
those interfaces even if firmware does not have support for them (yet...)
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-11-oohall@gmail.com
Currently we check that an IODA2 compatible PHB is upstream of this slot.
This is mainly to avoid pnv_php creating slots for the various "virtual
PHBs" that we create for NVLink. There's no real need for this restriction
so allow it on IODA3.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-10-oohall@gmail.com
When performing EEH recovery of devices in a hotplug slot we need to use
the slot driver's ->reset_slot() callback to prevent spurious hotplug
events due to spurious DLActive and PresDet change interrupts. Add a
reset_slot() callback to pnv_php so we can handle recovery of devices
in pnv_php managed slots.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-9-oohall@gmail.com
When we reset PCI devices managed by a hotplug driver the reset may
generate spurious hotplug events that cause the PCI device we're resetting
to be torn down accidently. This is a problem for EEH (when the driver is
EEH aware) since we want to leave the OS PCI device state intact so that
the device can be re-set without losing any resources (network, disks,
etc) provided by the driver.
Generic PCI code provides the pci_bus_error_reset() function to handle
resetting a PCI Device (or bus) by using the reset method provided by the
hotplug slot driver. We can use this function if the EEH core has
requested a hot reset (common case) without tripping over the hotplug
driver.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-8-oohall@gmail.com
Support for switching CAPI cards into and out of CAPI mode was removed a
while ago. Drop the comment since it's no longer relevant.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-7-oohall@gmail.com
Currently we print a stack trace in the event handler to help with
debugging EEH issues. In the case of suprise hot-unplug this is unneeded,
so we want to prevent printing the stack trace unless we know it's due to
an actual device error. To accomplish this, we can save a stack trace at
the point of detection and only print it once the EEH recovery handler has
determined the freeze was due to an actual error.
Since the whole point of this is to prevent spurious EEH output we also
move a few prints out of the detection thread, or mark them as pr_debug
so anyone interested can get output from the eeh_check_dev_failure()
if they want.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-6-oohall@gmail.com
When a device is surprise removed while undergoing IO we will probably
get an EEH PE freeze due to MMIO timeouts and other errors. When a freeze
is detected we send a recovery event to the EEH worker thread which will
notify drivers, and perform recovery as needed.
In the event of a hot-remove we don't want recovery to occur since there
isn't a device to recover. The recovery process is fairly long due to
the number of wait states (required by PCIe) which causes problems when
devices are removed and replaced (e.g. hot swapping of U.2 NVMe drives).
To determine if we need to skip the recovery process we can use the
get_adapter_state() operation of the hotplug_slot to determine if the
slot contains a device or not, and if the slot is empty we can skip
recovery entirely.
One thing to note is that the slot being EEH frozen does not prevent the
hotplug driver from working. We don't have the EEH recovery thread
remove any of the devices since it's assumed that the hotplug driver
will handle tearing down the slot state.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-5-oohall@gmail.com
If a device is torn down by a hotplug slot driver it's marked as removed
and marked as permaantly failed. There's no point in trying to recover a
permernantly failed device so it should be considered un-actionable.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-4-oohall@gmail.com
When hot-adding devices we rely on the hotplug driver to create pci_dn's
for the devices under the hotplug slot. Converse, when hot-removing the
driver will remove the pci_dn's that it created. This is a problem because
the pci_dev is still live until it's refcount drops to zero. This can
happen if the driver is slow to tear down it's internal state. Ideally, the
driver would not attempt to perform any config accesses to the device once
it's been marked as removed, but sometimes it happens. As a result, we
might attempt to access the pci_dn for a device that has been torn down and
the kernel may crash as a result.
To fix this, don't free the pci_dn unless the corresponding pci_dev has
been released. If the pci_dev is still live, then we mark the pci_dn with
a flag that indicates the pci_dev's release function should free it.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-3-oohall@gmail.com
When the last device in an eeh_pe is removed the eeh_pe structure itself
(and any empty parents) are freed since they are no longer needed. This
results in a crash when a hotplug driver is involved since the following
may occur:
1. Device is suprise removed.
2. Driver performs an MMIO, which fails and queues and eeh_event.
3. Hotplug driver receives a hotplug interrupt and removes any
pci_devs that were under the slot.
4. pci_dev is torn down and the eeh_pe is freed.
5. The EEH event handler thread processes the eeh_event and crashes
since the eeh_pe pointer in the eeh_event structure is no
longer valid.
Crashing is generally considered poor form. Instead of doing that use
the fact PEs are marked as EEH_PE_INVALID to keep them around until the
end of the recovery cycle, at which point we can safely prune any empty
PEs.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190903101605.2890-2-oohall@gmail.com
This avoids 3 loads in the radix page fault case, 1 load in the
hash fault case, and 2 loads in the hash miss page fault case.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190802105709.27696-37-npiggin@gmail.com
It is clever, but the small code saving is not worth the spaghetti of
jumping to a label in an expanded macro, particularly when the label
is just a number rather than a descriptive name.
So expand the INT_COMMON macro twice, once for the stack and no stack
cases, and branch to those. The slight code size increase is worth
the improved clarity of branches for this non-performance critical
code.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190802105709.27696-35-npiggin@gmail.com
This better reflects the order in which the code is executed.
No generated code change except BUG line number constants.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190802105709.27696-34-npiggin@gmail.com
Move DAR and DSISR saving to pt_regs into INT_COMMON. Also add an
option to expand RECONCILE_IRQ_STATE.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190802105709.27696-33-npiggin@gmail.com
Hari Bathini