Hi Russell,
Recently we tried to remove scsi_cmnd.tags struct member [0].
However it now shows that some of the arm SCSI drivers continue to use
this [1]. I think any other driver usage of this member had been found
and removed.
The impression is that the usage of scsi_cmnd.tag in those drivers is
quite dubious.
Now checking [2], it appears that you may have had some patches for
these drivers locally.
So is that the case? Is this HW still used with bleeding edge kernels?
If so, can we fix up this tag management?
[0]
https://lore.kernel.org/linux-scsi/[email protected]/T/#mb47909f38f35837686734369600051b278d124af
[1]
https://lore.kernel.org/linux-scsi/[email protected]/T/#md5d786e5753083b2f3e8e761b1c69809f82c7485
[2]
https://lore.kernel.org/lkml/[email protected]/
Thanks,
John
Hi Russell,
Have you had a chance to consider the below?
Thanks
>
> Recently we tried to remove scsi_cmnd.tags struct member [0].
>
> However it now shows that some of the arm SCSI drivers continue to use
> this [1]. I think any other driver usage of this member had been found
> and removed.
>
> The impression is that the usage of scsi_cmnd.tag in those drivers is
> quite dubious.
>
> Now checking [2], it appears that you may have had some patches for
> these drivers locally.
>
> So is that the case? Is this HW still used with bleeding edge kernels?
> If so, can we fix up this tag management?
>
> [0]
> https://lore.kernel.org/linux-scsi/[email protected]/T/#mb47909f38f35837686734369600051b278d124af
>
>
> [1]
> https://lore.kernel.org/linux-scsi/[email protected]/T/#md5d786e5753083b2f3e8e761b1c69809f82c7485
>
>
> [2]
> https://lore.kernel.org/lkml/[email protected]/
>
> Thanks,
> John
I haven't, sorry.
I have run 5.x kernels on the hardware, and do have a set of patches
kicking around for the SCSI drivers that do some cleanups. It looks
like the fixup is pretty simple from the links you've sent - using
scsi_cmd_to_rq() to get the tag.
That said, I think I may only had one SCSI drive that came anywhere
close to supported tagged queuing, so I never put much effort into
tagged command support. Both acornscsi and fas216 have it disabled
for this reason, so it's probably easier just to rip the tag code
out of these drivers.
Russell.
On Fri, Aug 27, 2021 at 03:55:05PM +0100, John Garry wrote:
> Hi Russell,
>
> Have you had a chance to consider the below?
>
> Thanks
>
>
> >
> > Recently we tried to remove scsi_cmnd.tags struct member [0].
> >
> > However it now shows that some of the arm SCSI drivers continue to use
> > this [1]. I think any other driver usage of this member had been found
> > and removed.
> >
> > The impression is that the usage of scsi_cmnd.tag in those drivers is
> > quite dubious.
> >
> > Now checking [2], it appears that you may have had some patches for
> > these drivers locally.
> >
> > So is that the case? Is this HW still used with bleeding edge kernels?
> > If so, can we fix up this tag management?
> >
> > [0] https://lore.kernel.org/linux-scsi/[email protected]/T/#mb47909f38f35837686734369600051b278d124af
> >
> >
> > [1] https://lore.kernel.org/linux-scsi/[email protected]/T/#md5d786e5753083b2f3e8e761b1c69809f82c7485
> >
> >
> > [2]
> > https://lore.kernel.org/lkml/[email protected]/
> >
> > Thanks,
> > John
>
>
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
On 8/27/21 5:09 PM, Russell King (Oracle) wrote:
> I haven't, sorry.
>
> I have run 5.x kernels on the hardware, and do have a set of patches
> kicking around for the SCSI drivers that do some cleanups. It looks
> like the fixup is pretty simple from the links you've sent - using
> scsi_cmd_to_rq() to get the tag.
>
> That said, I think I may only had one SCSI drive that came anywhere
> close to supported tagged queuing, so I never put much effort into
> tagged command support. Both acornscsi and fas216 have it disabled
> for this reason, so it's probably easier just to rip the tag code
> out of these drivers.
>
That's what I figured, too.
And that's what my patches do, killing the tag support from arm drivers
which had them disabled since the dawn of git history.
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
[email protected] +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
On 27/08/2021 16:23, Hannes Reinecke wrote:
> On 8/27/21 5:09 PM, Russell King (Oracle) wrote:
>> I haven't, sorry.
>>
>> I have run 5.x kernels on the hardware, and do have a set of patches
>> kicking around for the SCSI drivers that do some cleanups. It looks
>> like the fixup is pretty simple from the links you've sent - using
>> scsi_cmd_to_rq() to get the tag.
I'm not sure. The SCSI midlayer does nothing with that tag field, apart
from set it to the block layer request tag, and it seems that these
drivers reuse that field for their own tag management. But Hannes is
confident that we may just use scsi_cmd_to_rq().
>>
>> That said, I think I may only had one SCSI drive that came anywhere
>> close to supported tagged queuing, so I never put much effort into
>> tagged command support. Both acornscsi and fas216 have it disabled
>> for this reason, so it's probably easier just to rip the tag code
>> out of these drivers.
>>
> That's what I figured, too.
> And that's what my patches do, killing the tag support from arm drivers
> which had them disabled since the dawn of git history.
These are the patches:
https://lore.kernel.org/linux-scsi/[email protected]/T/#t
Hannes, please consider resending, cc'ing Russell.
Thanks,
John