Re: [PATCH v7 1/2] devcoredump: remove the useless gfp_t parameter in dev_coredumpv and dev_coredumpm

[duoming]

Hello,

On Thu, 18 Aug 2022 16:58:01 +0200 Greg KH wrote:

> On Wed, Aug 17, 2022 at 10:05:37PM +0800, duoming@xxxxxxxxxx wrote:
> > Hello,
> > 
> > On Wed, 17 Aug 2022 14:43:29 +0200 Greg KH wrote:
> > 
> > > On Wed, Aug 17, 2022 at 08:39:12PM +0800, Duoming Zhou wrote:
> > > > The dev_coredumpv() and dev_coredumpm() could not be used in atomic
> > > > context, because they call kvasprintf_const() and kstrdup() with
> > > > GFP_KERNEL parameter. The process is shown below:
> > > > 
> > > > dev_coredumpv(.., gfp_t gfp)
> > > >   dev_coredumpm(.., gfp_t gfp)
> > > >     dev_set_name
> > > >       kobject_set_name_vargs
> > > >         kvasprintf_const(GFP_KERNEL, ...); //may sleep
> > > >           kstrdup(s, GFP_KERNEL); //may sleep
> > > > 
> > > > This patch removes gfp_t parameter of dev_coredumpv() and dev_coredumpm()
> > > > and changes the gfp_t parameter of kzalloc() in dev_coredumpm() to
> > > > GFP_KERNEL in order to show they could not be used in atomic context.
> > > > 
> > > > Fixes: 833c95456a70 ("device coredump: add new device coredump class")
> > > > Reviewed-by: Brian Norris <briannorris@xxxxxxxxxxxx>
> > > > Reviewed-by: Johannes Berg <johannes@xxxxxxxxxxxxxxxx>
> > > > Signed-off-by: Duoming Zhou <duoming@xxxxxxxxxx>
> > > > ---
> > > > Changes in v7:
> > > >   - Remove gfp_t flag in amdgpu device.
> > > 
> > > Again, this creates a "flag day" where we have to be sure we hit all
> > > users of this api at the exact same time.  This will prevent any new
> > > driver that comes into a maintainer tree during the next 3 months from
> > > ever being able to use this api without cauing build breakages in the
> > > linux-next tree.
> > > 
> > > Please evolve this api to work properly for everyone at the same time,
> > > like was previously asked for so that we can take this change.  It will
> > > take 2 releases, but that's fine.
> > 
> > Thank you for your reply, I will evolve this api to work properly for everyone.
> > If there are not any new drivers that use this api during the next 3 months, 
> > I will send this patch again. Otherwise, I will wait until there are not new
> > users anymore.
> 
> No, that is not necessary.  Do the work now so that there is no flag day
> and you don't have to worry about new users, it will all "just work".

Do you mean we should replace dev_set_name() in dev_coredumpm() to some other
functions that could work both in interrupt context and process context?

Best regards,
Duoming Zhou