[PATCH] powerpc/eeh: trivial fix to non-conventional PCI address output on EEH log

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

Signed-off-by: Guilherme G. Piccoli <redacted>
---
 arch/powerpc/kernel/eeh.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/kernel/eeh.c b/arch/powerpc/kernel/eeh.c
index c9bc78e..7429556 100644
--- a/arch/powerpc/kernel/eeh.c
+++ b/arch/powerpc/kernel/eeh.c

@@ -168,10 +168,10 @@ static size_t eeh_dump_dev_log(struct eeh_dev *edev, char *buf, size_t len)
 	int n = 0, l = 0;
 	char buffer[128];
 
-	n += scnprintf(buf+n, len-n, "%04x:%02x:%02x:%01x\n",
+	n += scnprintf(buf+n, len-n, "%04x:%02x:%02x.%01x\n",
 		       edev->phb->global_number, pdn->busno,
 		       PCI_SLOT(pdn->devfn), PCI_FUNC(pdn->devfn));
-	pr_warn("EEH: of node=%04x:%02x:%02x:%01x\n",
+	pr_warn("EEH: of node=%04x:%02x:%02x.%01x\n",
 		edev->phb->global_number, pdn->busno,
 		PCI_SLOT(pdn->devfn), PCI_FUNC(pdn->devfn));
 

-- 
2.1.0

On Fri, Jul 22, 2016 at 02:05:29PM -0300, Guilherme G. Piccoli wrote:

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

Signed-off-by: Guilherme G. Piccoli <redacted>

Reviewed-by: Gavin Shan <redacted>

"Guilherme G. Piccoli" [off-list ref] writes:

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

What consumes the log? Can it cope with us changing the formatting?

cheers

On Mon, Jul 25, 2016 at 10:47:13AM +1000, Michael Ellerman wrote:

"Guilherme G. Piccoli" [off-list ref] writes:

quoted

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

What consumes the log? Can it cope with us changing the formatting?

The log is printed by pr_warn() as part of the EEH kernel log. Also,
it's argument passed to RTAS call "ibm,slot-error-detail" and it's
put into the user data section of the RTAS call's output, which is
used by RTAS daemon (rtasd) then. I don't see anyone expects fixed
format for it in the user data section.

The format was ever adjusted in commit 0ed352dddbfc ("powerpc/eeh:
Reduce lines of log dump") on Jul 17 2014. No complains received
against it so far. I guess nobody cares about the format or there
is a alarm isn't raised yet :)

Thanks,
Gavin

On 07/24/2016 10:46 PM, Gavin Shan wrote:

On Mon, Jul 25, 2016 at 10:47:13AM +1000, Michael Ellerman wrote:

quoted

"Guilherme G. Piccoli" [off-list ref] writes:

quoted

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

What consumes the log? Can it cope with us changing the formatting?

The log is printed by pr_warn() as part of the EEH kernel log. Also,
it's argument passed to RTAS call "ibm,slot-error-detail" and it's
put into the user data section of the RTAS call's output, which is
used by RTAS daemon (rtasd) then. I don't see anyone expects fixed
format for it in the user data section.

The format was ever adjusted in commit 0ed352dddbfc ("powerpc/eeh:
Reduce lines of log dump") on Jul 17 2014. No complains received
against it so far. I guess nobody cares about the format or there
is a alarm isn't raised yet :)

Thanks very much for the details Gavin.

You can "discard" my last reply then, since this explains the use of 
"ibm,slot-error-detail" correctly.

Cheers,


Guilherme

Thanks,
Gavin

_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

On 07/24/2016 10:46 PM, Gavin Shan wrote:

On Mon, Jul 25, 2016 at 10:47:13AM +1000, Michael Ellerman wrote:

quoted

"Guilherme G. Piccoli" [off-list ref] writes:

quoted

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

What consumes the log? Can it cope with us changing the formatting?

The log is printed by pr_warn() as part of the EEH kernel log. Also,
it's argument passed to RTAS call "ibm,slot-error-detail" and it's
put into the user data section of the RTAS call's output, which is
used by RTAS daemon (rtasd) then. I don't see anyone expects fixed
format for it in the user data section.

The format was ever adjusted in commit 0ed352dddbfc ("powerpc/eeh:
Reduce lines of log dump") on Jul 17 2014. No complains received
against it so far. I guess nobody cares about the format or there
is a alarm isn't raised yet :)

Thanks,
Gavin

Quick follow-up on this: RTAS daemon stores the information captured via 
ibm,slot-error-detail in a log file, which can be accessed using the 
command "rtas_dump -f /var/log/platform". More information on this can 
be found in 
https://www.ibm.com/support/knowledgecenter/linuxonibm/liaau/liaau-diagnosing-rtas-events.htm 
.

I was able to check this log and the EEH PCI address output was there, 
in ascii text format.

Thanks,

Guilherme

_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

On 07/24/2016 09:47 PM, Michael Ellerman wrote:

"Guilherme G. Piccoli" [off-list ref] writes:

quoted

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

What consumes the log? Can it cope with us changing the formatting?

Michael, as far as I understand (Russel and Gavin surely knows better 
and can correct me if I'm wrong) the only consumer that doesn't just 
output the error information to screen is pseries' get_log handler, 
which will pass this log information through a rtas_call 
(ibm_slot_error_detail).

Now, I imagine this information might be consumed in the end for the 
FSP, by reporting EEH errors there, but my testing in one LPAR showed me 
no errors in FSP after an EEH (without the patch applied). Maybe I'm 
missing a FSP configuration to be able to show the errors there?

Thanks,


Guilherme

cheers
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

On Fri, 2016-22-07 at 17:05:29 UTC, "Guilherme G. Piccoli" wrote:

This is a very minor/trivial fix for the output of PCI address on EEH logs.
The PCI address on "OF node" field currently is using ":" as a separator
for the function, but the usual separator is ".". This patch changes the
separator for dot, so the PCI address is printed as usual.

No functional changes were introduced.

Signed-off-by: Guilherme G. Piccoli <redacted>
Reviewed-by: Gavin Shan <redacted>

Applied to powerpc fixes, thanks.

https://git.kernel.org/powerpc/c/10560b9afc8abf349843dff88c

cheers