Re: [PATCH 5.10 1/1] io_uring: fix double free in the deferred/cancelled path

[Greg KH <greg@kroah.com>]

On Wed, Oct 27, 2021 at 10:03:01AM +0100, Lee Jones wrote:
> On Wed, 27 Oct 2021, Greg KH wrote:
> 
> > On Wed, Oct 27, 2021 at 09:37:59AM +0100, Lee Jones wrote:
> > > On Wed, 27 Oct 2021, Greg KH wrote:
> > > 
> > > > On Wed, Oct 27, 2021 at 09:01:28AM +0100, Lee Jones wrote:
> > > > > 792bb6eb86233 ("io_uring: don't take uring_lock during iowq cancel")
> > > > > inadvertently fixed this issue in v5.12.  This patch cherry-picks the
> > > > > hunk of commit which does so.
> > > > 
> > > > Why can't we take all of that commit?  Why only part of it?
> > > 
> > > I don't know.
> > > 
> > > Why didn't the Stable team take it further than v5.11.y?
> > 
> > Look in the archives?  Did it not apply cleanly?
> > 
> > /me goes off and looks...
> > 
> > Looks like I asked for a backport, but no one did it, I only received a
> > 5.11 version:
> > 	https://lore.kernel.org/r/1839646480a26a2461eccc38a75e98998d2d6e11.1615375332.git.asml.silence@gmail.com
> > 
> > so a 5.10 version would be nice, as I said it failed as-is:
> > 	https://lore.kernel.org/all/161460075611654@kroah.com/
> 
> Precisely.  This is the answer to your question:
> 
>   > > > Why can't we take all of that commit?  Why only part of it?
> 
> Same reason the Stable team didn't back-port it - it doesn't apply.
> 
> The second hunk is only relevant to v5.11+.

Great, then use the "normal" stable style, but down in the s-o-b area
say "dropped second chunk as it is not relevant to 5.10.y".

thanks,

greg k-h