Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Felipe Franciosi]

Hi Ming (and all),

Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automatic =
affinity" which forces virtio-scsi to use blk-mq fixes an issue introduced =
by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a72ad=
b), but your series was not backported to that stable branch.

Are there any plans to do that? At least CoreOS is using 4.14 and showing i=
ssues on AHV (which provides an mq virtio-scsi controller).

Thanks,
Felipe=

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Ming Lei]

On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
> Hi Ming (and all),
> 
> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automatic affinity" which forces virtio-scsi to use blk-mq fixes an issue introduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a72adb), but your series was not backported to that stable branch.
> 
> Are there any plans to do that? At least CoreOS is using 4.14 and showing issues on AHV (which provides an mq virtio-scsi controller).
> 

Hi Felipe,

Looks the following 4 patches should have been marked as stable, sorry
for missing that.

b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq vector affinity
2f31115e940c scsi: core: introduce force_blk_mq
adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
8b834bff1b73 scsi: hpsa: fix selection of reply queue

Usually this backporting is done by our stable guys, so I will CC stable
and leave them handle it, but I am happy to provide any help for
addressing conflicts or sort of thing.


Thanks,
Ming

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Greg Kroah-Hartman]

On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
> > Hi Ming (and all),
> > 
> > Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automatic affinity" which forces virtio-scsi to use blk-mq fixes an issue introduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a72adb), but your series was not backported to that stable branch.
> > 
> > Are there any plans to do that? At least CoreOS is using 4.14 and showing issues on AHV (which provides an mq virtio-scsi controller).
> > 
> 
> Hi Felipe,
> 
> Looks the following 4 patches should have been marked as stable, sorry
> for missing that.
> 
> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq vector affinity
> 2f31115e940c scsi: core: introduce force_blk_mq
> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
> 
> Usually this backporting is done by our stable guys, so I will CC stable
> and leave them handle it, but I am happy to provide any help for
> addressing conflicts or sort of thing.

As the above patches do not apply "cleanly" to the 4.14.y tree at all,
can you please provide a set of backported patches that I can apply?

thanks,

greg k-h

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Felipe Franciosi]

> On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation.org>=
 wrote:
>=20
> On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
>> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
>>> Hi Ming (and all),
>>>=20
>>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automa=
tic affinity" which forces virtio-scsi to use blk-mq fixes an issue introdu=
ced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a=
72adb), but your series was not backported to that stable branch.
>>>=20
>>> Are there any plans to do that? At least CoreOS is using 4.14 and showi=
ng issues on AHV (which provides an mq virtio-scsi controller).
>>>=20
>>=20
>> Hi Felipe,
>>=20
>> Looks the following 4 patches should have been marked as stable, sorry
>> for missing that.
>>=20
>> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq vect=
or affinity
>> 2f31115e940c scsi: core: introduce force_blk_mq
>> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
>> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
>>=20
>> Usually this backporting is done by our stable guys, so I will CC stable
>> and leave them handle it, but I am happy to provide any help for
>> addressing conflicts or sort of thing.
>=20
> As the above patches do not apply "cleanly" to the 4.14.y tree at all,
> can you please provide a set of backported patches that I can apply?

Actually, adbe552349f2 is already present in 4.14.y. It is commit e58114824=
fa6.

If you skip that, all the other three apply cleanly.

Thanks,
Felipe


>=20
> thanks,
>=20
> greg k-h

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Greg Kroah-Hartman]

On Fri, Aug 10, 2018 at 05:10:52PM +0000, Felipe Franciosi wrote:
> 
> > On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote:
> > 
> > On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
> >> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
> >>> Hi Ming (and all),
> >>> 
> >>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automatic affinity" which forces virtio-scsi to use blk-mq fixes an issue introduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a72adb), but your series was not backported to that stable branch.
> >>> 
> >>> Are there any plans to do that? At least CoreOS is using 4.14 and showing issues on AHV (which provides an mq virtio-scsi controller).
> >>> 
> >> 
> >> Hi Felipe,
> >> 
> >> Looks the following 4 patches should have been marked as stable, sorry
> >> for missing that.
> >> 
> >> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq vector affinity
> >> 2f31115e940c scsi: core: introduce force_blk_mq
> >> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
> >> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
> >> 
> >> Usually this backporting is done by our stable guys, so I will CC stable
> >> and leave them handle it, but I am happy to provide any help for
> >> addressing conflicts or sort of thing.
> > 
> > As the above patches do not apply "cleanly" to the 4.14.y tree at all,
> > can you please provide a set of backported patches that I can apply?
> 
> Actually, adbe552349f2 is already present in 4.14.y. It is commit e58114824fa6.
> 
> If you skip that, all the other three apply cleanly.

Ok, that works, but there's another bug report of aacraid having
problems.  Any ideas?

greg k-h

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Felipe Franciosi]

> On 10 Aug 2018, at 11:32, Greg Kroah-Hartman <gregkh@linuxfoundation.org>=
 wrote:
>=20
> On Fri, Aug 10, 2018 at 05:10:52PM +0000, Felipe Franciosi wrote:
>>=20
>>> On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation.or=
g> wrote:
>>>=20
>>> On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
>>>> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
>>>>> Hi Ming (and all),
>>>>>=20
>>>>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector auto=
matic affinity" which forces virtio-scsi to use blk-mq fixes an issue intro=
duced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f=
3a72adb), but your series was not backported to that stable branch.
>>>>>=20
>>>>> Are there any plans to do that? At least CoreOS is using 4.14 and sho=
wing issues on AHV (which provides an mq virtio-scsi controller).
>>>>>=20
>>>>=20
>>>> Hi Felipe,
>>>>=20
>>>> Looks the following 4 patches should have been marked as stable, sorry
>>>> for missing that.
>>>>=20
>>>> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq ve=
ctor affinity
>>>> 2f31115e940c scsi: core: introduce force_blk_mq
>>>> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
>>>> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
>>>>=20
>>>> Usually this backporting is done by our stable guys, so I will CC stab=
le
>>>> and leave them handle it, but I am happy to provide any help for
>>>> addressing conflicts or sort of thing.
>>>=20
>>> As the above patches do not apply "cleanly" to the 4.14.y tree at all,
>>> can you please provide a set of backported patches that I can apply?
>>=20
>> Actually, adbe552349f2 is already present in 4.14.y. It is commit e58114=
824fa6.
>>=20
>> If you skip that, all the other three apply cleanly.
>=20
> Ok, that works, but there's another bug report of aacraid having
> problems.  Any ideas?

Heya, I actually have no idea which bug you are talking about. TBH I'm only=
 experiencing the bug fixed by b5b6e8c8d3b4, which only requires 2f31115e94=
0c. (I tested a 4.14 with both commits which resolves the bug.)

I doubt any of that would have any interference with aacraid and should be =
safe backports in that respect.

Thanks,
Felipe

>=20
> greg k-h

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Felipe Franciosi]

> On 10 Aug 2018, at 20:43, Felipe Franciosi <felipe@nutanix.com> wrote:
>=20
>=20
>> On 10 Aug 2018, at 11:32, Greg Kroah-Hartman <gregkh@linuxfoundation.org=
> wrote:
>>=20
>> On Fri, Aug 10, 2018 at 05:10:52PM +0000, Felipe Franciosi wrote:
>>>=20
>>>> On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation.o=
rg> wrote:
>>>>=20
>>>> On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
>>>>> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
>>>>>> Hi Ming (and all),
>>>>>>=20
>>>>>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector aut=
omatic affinity" which forces virtio-scsi to use blk-mq fixes an issue intr=
oduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86=
f3a72adb), but your series was not backported to that stable branch.
>>>>>>=20
>>>>>> Are there any plans to do that? At least CoreOS is using 4.14 and sh=
owing issues on AHV (which provides an mq virtio-scsi controller).
>>>>>>=20
>>>>>=20
>>>>> Hi Felipe,
>>>>>=20
>>>>> Looks the following 4 patches should have been marked as stable, sorr=
y
>>>>> for missing that.
>>>>>=20
>>>>> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq v=
ector affinity
>>>>> 2f31115e940c scsi: core: introduce force_blk_mq
>>>>> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
>>>>> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
>>>>>=20
>>>>> Usually this backporting is done by our stable guys, so I will CC sta=
ble
>>>>> and leave them handle it, but I am happy to provide any help for
>>>>> addressing conflicts or sort of thing.
>>>>=20
>>>> As the above patches do not apply "cleanly" to the 4.14.y tree at all,
>>>> can you please provide a set of backported patches that I can apply?
>>>=20
>>> Actually, adbe552349f2 is already present in 4.14.y. It is commit e5811=
4824fa6.
>>>=20
>>> If you skip that, all the other three apply cleanly.
>>=20
>> Ok, that works, but there's another bug report of aacraid having
>> problems.  Any ideas?
>=20
> Heya, I actually have no idea which bug you are talking about. TBH I'm on=
ly experiencing the bug fixed by b5b6e8c8d3b4, which only requires 2f31115e=
940c. (I tested a 4.14 with both commits which resolves the bug.)
>=20
> I doubt any of that would have any interference with aacraid and should b=
e safe backports in that respect.

Hi Greg, sorry to bother but I didn't hear anything back about this. Are yo=
u picking up 2f31115e940c and b5b6e8c8d3b4 for 4.14.y or waiting for some o=
ther action?

If there's something I can do to help, please let me know. I've already tes=
ted a 4.14.62 with the two commits above and confirmed it fixes the problem=
 we're seeing.

Thanks,
Felipe=

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Greg Kroah-Hartman]

On Wed, Aug 15, 2018 at 02:30:23PM +0000, Felipe Franciosi wrote:
> 
> > On 10 Aug 2018, at 20:43, Felipe Franciosi <felipe@nutanix.com> wrote:
> > 
> > 
> >> On 10 Aug 2018, at 11:32, Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote:
> >> 
> >> On Fri, Aug 10, 2018 at 05:10:52PM +0000, Felipe Franciosi wrote:
> >>> 
> >>>> On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation.org> wrote:
> >>>> 
> >>>> On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
> >>>>> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
> >>>>>> Hi Ming (and all),
> >>>>>> 
> >>>>>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector automatic affinity" which forces virtio-scsi to use blk-mq fixes an issue introduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef86f3a72adb), but your series was not backported to that stable branch.
> >>>>>> 
> >>>>>> Are there any plans to do that? At least CoreOS is using 4.14 and showing issues on AHV (which provides an mq virtio-scsi controller).
> >>>>>> 
> >>>>> 
> >>>>> Hi Felipe,
> >>>>> 
> >>>>> Looks the following 4 patches should have been marked as stable, sorry
> >>>>> for missing that.
> >>>>> 
> >>>>> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq vector affinity
> >>>>> 2f31115e940c scsi: core: introduce force_blk_mq
> >>>>> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
> >>>>> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
> >>>>> 
> >>>>> Usually this backporting is done by our stable guys, so I will CC stable
> >>>>> and leave them handle it, but I am happy to provide any help for
> >>>>> addressing conflicts or sort of thing.
> >>>> 
> >>>> As the above patches do not apply "cleanly" to the 4.14.y tree at all,
> >>>> can you please provide a set of backported patches that I can apply?
> >>> 
> >>> Actually, adbe552349f2 is already present in 4.14.y. It is commit e58114824fa6.
> >>> 
> >>> If you skip that, all the other three apply cleanly.
> >> 
> >> Ok, that works, but there's another bug report of aacraid having
> >> problems.  Any ideas?
> > 
> > Heya, I actually have no idea which bug you are talking about. TBH I'm only experiencing the bug fixed by b5b6e8c8d3b4, which only requires 2f31115e940c. (I tested a 4.14 with both commits which resolves the bug.)
> > 
> > I doubt any of that would have any interference with aacraid and should be safe backports in that respect.
> 
> Hi Greg, sorry to bother but I didn't hear anything back about this.
> Are you picking up 2f31115e940c and b5b6e8c8d3b4 for 4.14.y or waiting
> for some other action?

They are in 4.14.63-rc1 right now, did you not see them?

thanks,

greg k-h

Re: Fix for 84676c1f (b5b6e8c8) missing in 4.14.y

[Felipe Franciosi]

> On 15 Aug 2018, at 16:28, Greg Kroah-Hartman <gregkh@linuxfoundation.org>=
 wrote:
>=20
> On Wed, Aug 15, 2018 at 02:30:23PM +0000, Felipe Franciosi wrote:
>>=20
>>> On 10 Aug 2018, at 20:43, Felipe Franciosi <felipe@nutanix.com> wrote:
>>>=20
>>>=20
>>>> On 10 Aug 2018, at 11:32, Greg Kroah-Hartman <gregkh@linuxfoundation.o=
rg> wrote:
>>>>=20
>>>> On Fri, Aug 10, 2018 at 05:10:52PM +0000, Felipe Franciosi wrote:
>>>>>=20
>>>>>> On 10 Aug 2018, at 03:15, Greg Kroah-Hartman <gregkh@linuxfoundation=
.org> wrote:
>>>>>>=20
>>>>>> On Fri, Aug 10, 2018 at 10:31:29AM +0800, Ming Lei wrote:
>>>>>>> On Fri, Aug 10, 2018 at 02:09:01AM +0000, Felipe Franciosi wrote:
>>>>>>>> Hi Ming (and all),
>>>>>>>>=20
>>>>>>>> Your series "scsi: virtio_scsi: fix IO hang caused by irq vector a=
utomatic affinity" which forces virtio-scsi to use blk-mq fixes an issue in=
troduced by 84676c1f. We noticed that this bug also exists in 4.14.y (as ef=
86f3a72adb), but your series was not backported to that stable branch.
>>>>>>>>=20
>>>>>>>> Are there any plans to do that? At least CoreOS is using 4.14 and =
showing issues on AHV (which provides an mq virtio-scsi controller).
>>>>>>>>=20
>>>>>>>=20
>>>>>>> Hi Felipe,
>>>>>>>=20
>>>>>>> Looks the following 4 patches should have been marked as stable, so=
rry
>>>>>>> for missing that.
>>>>>>>=20
>>>>>>> b5b6e8c8d3b4 scsi: virtio_scsi: fix IO hang caused by automatic irq=
 vector affinity
>>>>>>> 2f31115e940c scsi: core: introduce force_blk_mq
>>>>>>> adbe552349f2 scsi: megaraid_sas: fix selection of reply queue
>>>>>>> 8b834bff1b73 scsi: hpsa: fix selection of reply queue
>>>>>>>=20
>>>>>>> Usually this backporting is done by our stable guys, so I will CC s=
table
>>>>>>> and leave them handle it, but I am happy to provide any help for
>>>>>>> addressing conflicts or sort of thing.
>>>>>>=20
>>>>>> As the above patches do not apply "cleanly" to the 4.14.y tree at al=
l,
>>>>>> can you please provide a set of backported patches that I can apply?
>>>>>=20
>>>>> Actually, adbe552349f2 is already present in 4.14.y. It is commit e58=
114824fa6.
>>>>>=20
>>>>> If you skip that, all the other three apply cleanly.
>>>>=20
>>>> Ok, that works, but there's another bug report of aacraid having
>>>> problems.  Any ideas?
>>>=20
>>> Heya, I actually have no idea which bug you are talking about. TBH I'm =
only experiencing the bug fixed by b5b6e8c8d3b4, which only requires 2f3111=
5e940c. (I tested a 4.14 with both commits which resolves the bug.)
>>>=20
>>> I doubt any of that would have any interference with aacraid and should=
 be safe backports in that respect.
>>=20
>> Hi Greg, sorry to bother but I didn't hear anything back about this.
>> Are you picking up 2f31115e940c and b5b6e8c8d3b4 for 4.14.y or waiting
>> for some other action?
>=20
> They are in 4.14.63-rc1 right now, did you not see them?

Hi Greg, terribly sorry. I fetched the wrong remote and didn't see them bef=
ore. Double checked now and it's there. :)

Thanks!
F.

>=20
> thanks,
>=20
> greg k-h