[PATCH] PCI/AER: Add a null check before eeh_ops->notify_resume callback.

This patch puts a NULL check before branching to the address pointed
to by eeh_ops->notify_resume in eeh_report_resume(). The callback
is used to notify the arch EEH code that a pci device is back
online.

For PPC64 presently, only an implementation for pseries platform is
available and not for powernv. Hence without this patch EEH recovery
on all non-virtualized hosts is causing a kernel panic when
CONFIG_PCI_IOV is set. The panic is usually is of the form:

EEH: Notify device driver to resume
Unable to handle kernel paging request for instruction fetch
Faulting instruction address: 0x00000000
Oops: Kernel access of bad area, sig: 11 [#1]
<snip>
LR eeh_report_resume+0x218/0x220
Call Trace:
 eeh_report_resume+0x1f0/0x220 (unreliable)
 eeh_pe_dev_traverse+0x98/0x170
 eeh_handle_normal_event+0x3f4/0x650
 eeh_handle_event+0x188/0x380
 eeh_event_handler+0x208/0x210
 kthread+0x168/0x1b0
 ret_from_kernel_thread+0x5c/0xb4

Cc: Bryant G. Ly <redacted>
Fixes: 856e1eb9bdd4("PCI/AER: Add uevents in AER and EEH error/resume")
Signed-off-by: Vaibhav Jain <redacted>
---
 arch/powerpc/kernel/eeh_driver.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/kernel/eeh_driver.c b/arch/powerpc/kernel/eeh_driver.c
index beea2182d754..932858a293ea 100644
--- a/arch/powerpc/kernel/eeh_driver.c
+++ b/arch/powerpc/kernel/eeh_driver.c

@@ -384,7 +384,8 @@ static void *eeh_report_resume(void *data, void *userdata)
 	eeh_pcid_put(dev);
 	pci_uevent_ers(dev, PCI_ERS_RESULT_RECOVERED);
 #ifdef CONFIG_PCI_IOV
-	eeh_ops->notify_resume(eeh_dev_to_pdn(edev));
+	if (eeh_ops->notify_resume)
+		eeh_ops->notify_resume(eeh_dev_to_pdn(edev));
 #endif
 	return NULL;
 }

-- 
2.14.3

There is already a patch for this issue applied to ppc-next viz commit
521ca5a9859a870e354d1a6b84a6ff ("powerpc/eeh: Add conditional check on
notify_resume"). So please ignore the patch.

-- 
Vaibhav Jain [off-list ref]
Linux Technology Center, IBM India Pvt. Ltd.

On 2/22/18 5:58 AM, Vaibhav Jain wrote:

A version of this patch already upstreamed. 

https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?h=fixes&id=521ca5a9859a870e354d1a6b84a6ff4c07bbceb0

-Bryant

Vaibhav Jain [off-list ref] writes:

This patch puts a NULL check before branching to the address pointed
to by eeh_ops->notify_resume in eeh_report_resume(). The callback
is used to notify the arch EEH code that a pci device is back
online.

For PPC64 presently, only an implementation for pseries platform is
available and not for powernv. Hence without this patch EEH recovery
on all non-virtualized hosts is causing a kernel panic when
CONFIG_PCI_IOV is set. The panic is usually is of the form:

EEH: Notify device driver to resume
Unable to handle kernel paging request for instruction fetch
Faulting instruction address: 0x00000000
Oops: Kernel access of bad area, sig: 11 [#1]
<snip>
LR eeh_report_resume+0x218/0x220
Call Trace:
 eeh_report_resume+0x1f0/0x220 (unreliable)
 eeh_pe_dev_traverse+0x98/0x170
 eeh_handle_normal_event+0x3f4/0x650
 eeh_handle_event+0x188/0x380
 eeh_event_handler+0x208/0x210
 kthread+0x168/0x1b0
 ret_from_kernel_thread+0x5c/0xb4

Cc: Bryant G. Ly <redacted>
Fixes: 856e1eb9bdd4("PCI/AER: Add uevents in AER and EEH error/resume")
Signed-off-by: Vaibhav Jain <redacted>

10 out of 10 for the change log!

But yeah this is already fixed in my fixes branch, thanks anyway.

cheers