Backport request: nfsd: Fix NFSv4 READ on RDMA when using readv

Backport request: nfsd: Fix NFSv4 READ on RDMA when using readv

[Timo Rothenpieler]

Sorry if this is not the right approach to this, but I'd like to request 
a backport of 412055398b9e67e07347a936fc4a6adddabe9cf4, "nfsd: Fix NFSv4 
READ on RDMA when using readv" to Linux 5.4.

The patch applies cleanly and fixes a rare but severe issue with NFS 
over RDMA, which I just spent several days tracking down to that patch, 
with major help from linux-nfs/rdma.

I have right now manually applied it to my 5.4 Kernel and can confirm it 
fixes the issue.



Thanks,
Timo

Re: Backport request: nfsd: Fix NFSv4 READ on RDMA when using readv

[Greg KH]

On Tue, Aug 04, 2020 at 10:47:26PM +0200, Timo Rothenpieler wrote:
> Sorry if this is not the right approach to this, but I'd like to request a
> backport of 412055398b9e67e07347a936fc4a6adddabe9cf4, "nfsd: Fix NFSv4 READ
> on RDMA when using readv" to Linux 5.4.
> 
> The patch applies cleanly and fixes a rare but severe issue with NFS over
> RDMA, which I just spent several days tracking down to that patch, with
> major help from linux-nfs/rdma.
> 
> I have right now manually applied it to my 5.4 Kernel and can confirm it
> fixes the issue.

It's good to cc: the developers/maintainers of the patch you are asking
about, to see if they object or not.

Chuck, any objection for the above patch being added to 5.4 and maybe
4.19?

thanks,

greg k-h

Re: Backport request: nfsd: Fix NFSv4 READ on RDMA when using readv

[Chuck Lever]

Hi Greg-

> On Aug 10, 2020, at 9:52 AM, Greg KH <gregkh@linuxfoundation.org> wrote:
> 
> On Tue, Aug 04, 2020 at 10:47:26PM +0200, Timo Rothenpieler wrote:
>> Sorry if this is not the right approach to this, but I'd like to request a
>> backport of 412055398b9e67e07347a936fc4a6adddabe9cf4, "nfsd: Fix NFSv4 READ
>> on RDMA when using readv" to Linux 5.4.
>> 
>> The patch applies cleanly and fixes a rare but severe issue with NFS over
>> RDMA, which I just spent several days tracking down to that patch, with
>> major help from linux-nfs/rdma.
>> 
>> I have right now manually applied it to my 5.4 Kernel and can confirm it
>> fixes the issue.
> 
> It's good to cc: the developers/maintainers of the patch you are asking
> about, to see if they object or not.
> 
> Chuck, any objection for the above patch being added to 5.4 and maybe
> 4.19?

Since Timo applied it to 5.4 and tested it, I have no objection there.

v4.19 suffers from the same bug, but I don't have reports of anyone
having tried 4120553 on that kernel. I would say wait on that one until
someone can state that it works without regression.


--
Chuck Lever

Re: Backport request: nfsd: Fix NFSv4 READ on RDMA when using readv

[Greg KH]

On Mon, Aug 10, 2020 at 09:57:56AM -0400, Chuck Lever wrote:
> Hi Greg-
> 
> > On Aug 10, 2020, at 9:52 AM, Greg KH <gregkh@linuxfoundation.org> wrote:
> > 
> > On Tue, Aug 04, 2020 at 10:47:26PM +0200, Timo Rothenpieler wrote:
> >> Sorry if this is not the right approach to this, but I'd like to request a
> >> backport of 412055398b9e67e07347a936fc4a6adddabe9cf4, "nfsd: Fix NFSv4 READ
> >> on RDMA when using readv" to Linux 5.4.
> >> 
> >> The patch applies cleanly and fixes a rare but severe issue with NFS over
> >> RDMA, which I just spent several days tracking down to that patch, with
> >> major help from linux-nfs/rdma.
> >> 
> >> I have right now manually applied it to my 5.4 Kernel and can confirm it
> >> fixes the issue.
> > 
> > It's good to cc: the developers/maintainers of the patch you are asking
> > about, to see if they object or not.
> > 
> > Chuck, any objection for the above patch being added to 5.4 and maybe
> > 4.19?
> 
> Since Timo applied it to 5.4 and tested it, I have no objection there.
> 
> v4.19 suffers from the same bug, but I don't have reports of anyone
> having tried 4120553 on that kernel. I would say wait on that one until
> someone can state that it works without regression.

Ok, sounds good, now queued up for 5.4.y, thanks!

greg k-h