All of lore.kernel.org
 help / color / mirror / Atom feed
* rdma-core: pvrdma init issue
@ 2017-01-19 14:05 Yishai Hadas
       [not found] ` <b13a0a91-f93a-1072-0e5a-e9841660cf00-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
  0 siblings, 1 reply; 3+ messages in thread
From: Yishai Hadas @ 2017-01-19 14:05 UTC (permalink / raw)
  To: ditr-pghWNbHTmq7QT0dZR+AlfA
  Cc: yishaih-VPRAkNaXOzVWk0Htik3J/w,
	linux-rdma (linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org),
	Doug Ledford, Jason Gunthorpe, liranl-VPRAkNaXOzVWk0Htik3J/w,
	Majd Dibbiny, Leon Romanovsky

Hi Adit,

Working with latest rdma-core stuff that includes the pvrdma provider, 
encountered below issue, would appreciate your input on.

While libibverbs loads the drivers there are few "annoying" warnings 
that are printed from pvrdma per uverbs sysfs entry coming from 
'pvrdma_driver_init_shared'.

"pvrdma: ABI version 4 of /sys/class/infiniband_verbs/uverbs0 is not 
supported (supported 3)
pvrdma: ABI version 1 of /sys/class/infiniband_verbs/uverbs1 is not 
supported (supported 3)
...
"

Similar messages basically appear in other drivers after some pre- 
detection that the sysfs entry relates to their driver based on vendor 
id, device name but the ABI is not supported.(see mlx5_driver_init)

In the pvrdma driver you are using the ABI version as the detection and 
as a result print some redundant/annoying messages into stdout when 
working with some other device. (e.g. ibv_rc_pingpong -d mlx5_0).

In addition,
Can you please explain the detection logic itself ? your code chose the 
value of 3 (i.e. PVRDMA_UVERBS_ABI_VERSION) with some comment as of:
"We support only a single ABI version for now".

What if some other kernel driver uses that value as well or in the 
future you'll add some value that is already in use ? It might cause 
that pvrdma user driver will be used for that device without a real plan 
by the application.

For example if mlx4_0 will report same ABI version the call will end-up 
with an ib_device having below information:
"dev_name=uverbs0, drive_name=pvrdma, name=mlx4_0"

Yishai
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: rdma-core: pvrdma init issue
       [not found] ` <b13a0a91-f93a-1072-0e5a-e9841660cf00-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
@ 2017-01-19 14:48   ` Leon Romanovsky
  2017-01-19 16:27   ` Adit Ranadive
  1 sibling, 0 replies; 3+ messages in thread
From: Leon Romanovsky @ 2017-01-19 14:48 UTC (permalink / raw)
  To: aditr-pghWNbHTmq7QT0dZR+AlfA
  Cc: Yishai Hadas,
	linux-rdma (linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org),
	Doug Ledford, Jason Gunthorpe, liranl-VPRAkNaXOzVWk0Htik3J/w,
	Majd Dibbiny

[-- Attachment #1: Type: text/plain, Size: 1910 bytes --]

And now with correct email address.

On Thu, Jan 19, 2017 at 04:05:36PM +0200, Yishai Hadas wrote:
> Hi Adit,
>
> Working with latest rdma-core stuff that includes the pvrdma provider,
> encountered below issue, would appreciate your input on.
>
> While libibverbs loads the drivers there are few "annoying" warnings that
> are printed from pvrdma per uverbs sysfs entry coming from
> 'pvrdma_driver_init_shared'.
>
> "pvrdma: ABI version 4 of /sys/class/infiniband_verbs/uverbs0 is not
> supported (supported 3)
> pvrdma: ABI version 1 of /sys/class/infiniband_verbs/uverbs1 is not
> supported (supported 3)
> ...
> "
>
> Similar messages basically appear in other drivers after some pre- detection
> that the sysfs entry relates to their driver based on vendor id, device name
> but the ABI is not supported.(see mlx5_driver_init)
>
> In the pvrdma driver you are using the ABI version as the detection and as a
> result print some redundant/annoying messages into stdout when working with
> some other device. (e.g. ibv_rc_pingpong -d mlx5_0).
>
> In addition,
> Can you please explain the detection logic itself ? your code chose the
> value of 3 (i.e. PVRDMA_UVERBS_ABI_VERSION) with some comment as of:
> "We support only a single ABI version for now".
>
> What if some other kernel driver uses that value as well or in the future
> you'll add some value that is already in use ? It might cause that pvrdma
> user driver will be used for that device without a real plan by the
> application.
>
> For example if mlx4_0 will report same ABI version the call will end-up with
> an ib_device having below information:
> "dev_name=uverbs0, drive_name=pvrdma, name=mlx4_0"
>
> Yishai
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: rdma-core: pvrdma init issue
       [not found] ` <b13a0a91-f93a-1072-0e5a-e9841660cf00-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
  2017-01-19 14:48   ` Leon Romanovsky
@ 2017-01-19 16:27   ` Adit Ranadive
  1 sibling, 0 replies; 3+ messages in thread
From: Adit Ranadive @ 2017-01-19 16:27 UTC (permalink / raw)
  To: Yishai Hadas
  Cc: yishaih-VPRAkNaXOzVWk0Htik3J/w,
	linux-rdma (linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org),
	Doug Ledford, Jason Gunthorpe, liranl-VPRAkNaXOzVWk0Htik3J/w,
	Majd Dibbiny, Leon Romanovsky

Hi Yishai,

On Thu, Jan 19, 2017 at 4:05:36PM +0200, Yishai Hadas wrote:
> Hi Adit,
> 
> Working with latest rdma-core stuff that includes the pvrdma provider, 
> encountered below issue, would appreciate your input on.
> 
> While libibverbs loads the drivers there are few "annoying" warnings 
> that are printed from pvrdma per uverbs sysfs entry coming from 
> 'pvrdma_driver_init_shared'.
> 
> "pvrdma: ABI version 4 of /sys/class/infiniband_verbs/uverbs0 is not 
> supported (supported 3)
> pvrdma: ABI version 1 of /sys/class/infiniband_verbs/uverbs1 is not 
> supported (supported 3)
> ...
> "

We have a fix that addresses this internally. We will have to check for
the device and vendor id before the ABI check. I will send that out shortly.

Thanks,
Adit

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2017-01-19 16:27 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-01-19 14:05 rdma-core: pvrdma init issue Yishai Hadas
     [not found] ` <b13a0a91-f93a-1072-0e5a-e9841660cf00-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2017-01-19 14:48   ` Leon Romanovsky
2017-01-19 16:27   ` Adit Ranadive

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.