Re: [PATCH v7] devcoredump: change gfp_t parameter of kzalloc to GFP_KERNEL

[duoming@zju.edu.cn]

Hello,

On Tue, 28 Jun 2022 14:18:29 +0200 greg KH 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)
> > > >     kzalloc(.., gfp);
> > > >     dev_set_name
> > > >       kobject_set_name_vargs
> > > >         kvasprintf_const(GFP_KERNEL, ...); //may sleep
> > > >           kstrdup(s, GFP_KERNEL); //may sleep
> > > > 
> > > > This patch 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.
> > > > 
> > > > What's more, this patch does not remove the gfp_t parameter in
> > > > dev_coredumpv() and dev_coredumpm() in order that it will not influence
> > > > other new users that are added in other trees.
> > > > 
> > > > Fixes: 833c95456a70 ("device coredump: add new device coredump class")
> > > > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > > > ---
> > > > Changes in v7:
> > > >   - change gfp_t parameter of kzalloc in dev_coredumpm() to GFP_KERNEL.
> > > > 
> > > >  drivers/base/devcoredump.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c
> > > > index f4d794d6bb8..cf60aacf8a8 100644
> > > > --- a/drivers/base/devcoredump.c
> > > > +++ b/drivers/base/devcoredump.c
> > > > @@ -268,7 +268,7 @@ void dev_coredumpm(struct device *dev, struct module *owner,
> > > >  	if (!try_module_get(owner))
> > > >  		goto free;
> > > >  
> > > > -	devcd = kzalloc(sizeof(*devcd), gfp);
> > > > +	devcd = kzalloc(sizeof(*devcd), GFP_KERNEL);
> > > 
> > > No, you can't just ignore the flag entirely, that doesn't help anyone
> > > out who tries to set it and is totally confused as to why the field is
> > > ignored.
> > > 
> > > You need to evolve the function over time to not need the parameter at
> > > all, this just papers over the entire issue, which makes the api lie to
> > > the caller, not something you ever want to do.
> > 
> > Thank you for your time and reply.
> > 
> > But if there are new devices come into kernel, it may use devcoredump api.
> > What is the proper time to remove the gfp_t parameter of dev_coredumpv()
> > and dev_coredumpm()?
> 
> Normally you prepare some patches that does the conversion as a patch
> series and I queue them up in my tree, and get them merged in -rc1, then
> any stragglers are then fixed up in -rc2 along with the final rename of
> the old way and then all is good.  See lots of examples of changing apis
> over time on the mailing lists for how to do this.

Thank you for your time and reply, I understand.

I will resend the patch that removes the gfp_t parameter of dev_coredumpv() and dev_coredumpm()
until commit oc3d8785f6c04a ("drm/amdgpu: adding device coredump support") has been merged into
mainline. Maybe after v5.19-rc5 or v5.19-rc6.

Best regards,
Duoming Zhou