Skip to contentSkip to conversation

libndctl: Fix probe of non-nfit nvdimms

PATCH In review Vaibhav Jain 4 participants 5 messages Oct 9, 2020 – Oct 12, 2020 libnvdimm: non-volatile memory device

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libndctl: Fix probe of non-nfit nvdimms

commit 107a24ff429f ("ndctl/list: Add firmware activation enumeration") introduced changes in add_dimm() to enumerate the status of firmware activation. However a branch added in that commit broke the probe for non-nfit nvdimms like one provided by papr-scm. This cause an error reported when listing namespaces like below:

$ sudo ndctl list
libndctl: add_dimm: nmem0: probe failed: No such device
libndctl: __sysfs_device_parse: nmem0: add_dev() failed

Do a fix for this by removing the offending branch in the add_dimm() patch. This continues the flow of add_dimm() probe even if the nfit is not detected on the associated bus.

Fixes 107a24ff429fa("ndctl/list: Add firmware activation enumeration")
---
 ndctl/lib/libndctl.c | 3 ---
 1 file changed, 3 deletions(-)
ndctl/lib/libndctl.c +0 -3
index 554696386f48..ad521d365ee4 100644
@@ -1875,9 +1875,6 @@ static void *add_dimm(void *parent, int id, const char *dimm_base)
 	else
 		dimm->fwa_result = fwa_result_to_result(buf);
 
-	if (!ndctl_bus_has_nfit(bus))
-		goto out;
-
 	/* Check if the given dimm supports nfit */
 	if (ndctl_bus_has_nfit(bus)) {
 		dimm->formats = formats;
Signature · 4 lines
2.26.2
_______________________________________________
Linux-nvdimm mailing list -- [email protected]
To unsubscribe send an email to [email protected]
maintainer
#1 lore raw
Quoted 14 lines · Vaibhav Jain

commit 107a24ff429f ("ndctl/list: Add firmware activation enumeration") introduced changes in add_dimm() to enumerate the status of firmware activation. However a branch added in that commit broke the probe for non-nfit nvdimms like one provided by papr-scm. This cause an error reported when listing namespaces like below:

$ sudo ndctl list
libndctl: add_dimm: nmem0: probe failed: No such device
libndctl: __sysfs_device_parse: nmem0: add_dev() failed

Do a fix for this by removing the offending branch in the add_dimm() patch. This continues the flow of add_dimm() probe even if the nfit is not detected on the associated bus.

Quoted 5 lines
Fixes 107a24ff429fa("ndctl/list: Add firmware activation enumeration")
---
 ndctl/lib/libndctl.c | 3 ---
 1 file changed, 3 deletions(-)

Ah apologies - this snuck in when I reflowed Dan's patches on top of the papr work for v70.

I expect you'd like a point release with this fix asap?

Is there a way for me to incorporate some papr unit tests into my release workflow so I can avoid breaking things like this again?

I'll also try to do a better job of pushing things out to the pending branch more frequently so if you're monitoring that branch, hopefully things like this will get caught before a release happens :)

Quoted 6 lines
ndctl/lib/libndctl.c +0 -0
index 554696386f48..ad521d365ee4 100644

ndctl/lib/libndctl.c static void *add_dimm(void *parent, int id, const char *dimm_base) · line 1880

Quoted 9 lines
ndctl/lib/libndctl.c +0 -3
@@ -1875,9 +1875,6 @@ static void *add_dimm(void *parent, int id, const char *dimm_base)
 	else
 		dimm->fwa_result = fwa_result_to_result(buf);
 
-	if (!ndctl_bus_has_nfit(bus))
-		goto out;
-
 	/* Check if the given dimm supports nfit */
 	if (ndctl_bus_has_nfit(bus)) {
 		dimm->formats = formats;

_______________________________________________
Linux-nvdimm mailing list -- [email protected] To unsubscribe send an email to [email protected]

#2 lore raw
Quoted 25 lines · Vishal L
Quoted 14 lines · Vaibhav Jain

commit 107a24ff429f ("ndctl/list: Add firmware activation enumeration") introduced changes in add_dimm() to enumerate the status of firmware activation. However a branch added in that commit broke the probe for non-nfit nvdimms like one provided by papr-scm. This cause an error reported when listing namespaces like below:

$ sudo ndctl list
libndctl: add_dimm: nmem0: probe failed: No such device
libndctl: __sysfs_device_parse: nmem0: add_dev() failed

Do a fix for this by removing the offending branch in the add_dimm() patch. This continues the flow of add_dimm() probe even if the nfit is not detected on the associated bus.

Quoted 5 lines
Fixes 107a24ff429fa("ndctl/list: Add firmware activation enumeration")
---
 ndctl/lib/libndctl.c | 3 ---
 1 file changed, 3 deletions(-)

Ah apologies - this snuck in when I reflowed Dan's patches on top of the papr work for v70.

Quoted 8 lines

I expect you'd like a point release with this fix asap?

Is there a way for me to incorporate some papr unit tests into my release workflow so I can avoid breaking things like this again?

I'll also try to do a better job of pushing things out to the pending branch more frequently so if you're monitoring that branch, hopefully things like this will get caught before a release happens :)

Would be nice to have something like a papr_test next to nfit_test for such regression testing. These kinds of mistakes are really only avoidable with regression tests.
_______________________________________________
Linux-nvdimm mailing list -- [email protected]
To unsubscribe send an email to [email protected]

#3 lore raw

Hi Dan and Vishal,

Thanks so much for quick turnaround on this.

Quoted 18 lines · Dan Williams
Quoted 16 lines · Vishal L
Quoted 14 lines · Vaibhav Jain

commit 107a24ff429f ("ndctl/list: Add firmware activation enumeration") introduced changes in add_dimm() to enumerate the status of firmware activation. However a branch added in that commit broke the probe for non-nfit nvdimms like one provided by papr-scm. This cause an error reported when listing namespaces like below:

$ sudo ndctl list
libndctl: add_dimm: nmem0: probe failed: No such device
libndctl: __sysfs_device_parse: nmem0: add_dev() failed

Do a fix for this by removing the offending branch in the add_dimm() patch. This continues the flow of add_dimm() probe even if the nfit is not detected on the associated bus.

Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
Fixes 107a24ff429fa("ndctl/list: Add firmware activation enumeration")
---
 ndctl/lib/libndctl.c | 3 ---
 1 file changed, 3 deletions(-)

Ah apologies - this snuck in when I reflowed Dan's patches on top of the papr work for v70.

No worries :-)

Quoted 2 lines
Quoted 2 lines

I expect you'd like a point release with this fix asap?

Yes, that will be great. Thanks

Quoted 7 lines
Quoted 7 lines

Is there a way for me to incorporate some papr unit tests into my release workflow so I can avoid breaking things like this again?

I'll also try to do a better job of pushing things out to the pending branch more frequently so if you're monitoring that branch, hopefully things like this will get caught before a release happens :)

Fully agree, if that happens we can incorporate it into our CI system to ensure that such regressions are caught early on before any release is tagged.

Quoted 4 lines

Would be nice to have something like a papr_test next to nfit_test for such regression testing. These kinds of mistakes are really only avoidable with regression tests.

Yes Agree, fortunatly Santosh has recently posted an RFC patchset implementing such tests at [1]. Once that gets merged, can used to perform regression testing.

[1] "testing/nvdimm: Add test module for non-nfit platforms"
https://lore.kernel.org/linux-nvdimm/[email protected]/

Signature · 5 lines
Cheers
~ Vaibhav
_______________________________________________
Linux-nvdimm mailing list -- [email protected]
To unsubscribe send an email to [email protected]
#4 lore raw
Quoted 54 lines · Vaibhav Jain

Hi Dan and Vishal,

Thanks so much for quick turnaround on this.

Quoted 18 lines · Dan Williams
Quoted 16 lines · Vishal L
Quoted 14 lines · Vaibhav Jain

commit 107a24ff429f ("ndctl/list: Add firmware activation enumeration") introduced changes in add_dimm() to enumerate the status of firmware activation. However a branch added in that commit broke the probe for non-nfit nvdimms like one provided by papr-scm. This cause an error reported when listing namespaces like below:

$ sudo ndctl list
libndctl: add_dimm: nmem0: probe failed: No such device
libndctl: __sysfs_device_parse: nmem0: add_dev() failed

Do a fix for this by removing the offending branch in the add_dimm() patch. This continues the flow of add_dimm() probe even if the nfit is not detected on the associated bus.

Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
Fixes 107a24ff429fa("ndctl/list: Add firmware activation enumeration")
---
 ndctl/lib/libndctl.c | 3 ---
 1 file changed, 3 deletions(-)

Ah apologies - this snuck in when I reflowed Dan's patches on top of the papr work for v70.

No worries :-)

Quoted 2 lines
Quoted 2 lines

I expect you'd like a point release with this fix asap?

Yes, that will be great. Thanks

Quoted 7 lines
Quoted 7 lines

Is there a way for me to incorporate some papr unit tests into my release workflow so I can avoid breaking things like this again?

I'll also try to do a better job of pushing things out to the pending branch more frequently so if you're monitoring that branch, hopefully things like this will get caught before a release happens :)

Fully agree, if that happens we can incorporate it into our CI system to ensure that such regressions are caught early on before any release is tagged.

Quoted 4 lines

Would be nice to have something like a papr_test next to nfit_test for such regression testing. These kinds of mistakes are really only avoidable with regression tests.

Quoted 7 lines

Yes Agree, fortunatly Santosh has recently posted an RFC patchset implementing such tests at [1]. Once that gets merged, can used to perform regression testing.

[1] "testing/nvdimm: Add test module for non-nfit platforms"
https://lore.kernel.org/linux-nvdimm/[email protected]/

Thanks Vaibhav to point that out.

Dan/Vishal/Ira,

If you could provide your comments on the above RFC we could move forward on this.

Thanks,
Santosh
_______________________________________________
Linux-nvdimm mailing list -- [email protected] To unsubscribe send an email to [email protected]