On Mon, Sep 26, 2022 at 02:16:09PM +0800, Duoming Zhou wrote:
> The dev_coredumpm(), dev_coredumpv() and dev_coredumpsg() 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 adds context check in dev_coredumpm() in order to show
> dev_coredumpm() and its callers could not be used in atomic context.
>
> What's more, this change can allow the api to evolve and will not
> influence the users that call this api.
>
> Fixes: 833c95456a70 ("device coredump: add new device coredump class")
> Signed-off-by: Duoming Zhou <[email protected]>
> ---
> Changes in v9:
> - Add context check in dev_coredumpm().
>
> drivers/base/devcoredump.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c
> index f4d794d6bb8..806ee872f5f 100644
> --- a/drivers/base/devcoredump.c
> +++ b/drivers/base/devcoredump.c
> @@ -255,6 +255,9 @@ void dev_coredumpm(struct device *dev, struct module *owner,
> struct devcd_entry *devcd;
> struct device *existing;
>
> + if (!gfpflags_normal_context(gfp))
> + goto free;
Wait, this just broke things for no good reason if a caller happened to
have set a GFP flag that we do not like at the moment.
What code now does not work that previously did work with this change?
thanks,
greg k-h
Hello,
On Mon, 26 Sep 2022 08:27:58 +0200 Greg KH wrote:
> On Mon, Sep 26, 2022 at 02:16:09PM +0800, Duoming Zhou wrote:
> > The dev_coredumpm(), dev_coredumpv() and dev_coredumpsg() 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 adds context check in dev_coredumpm() in order to show
> > dev_coredumpm() and its callers could not be used in atomic context.
> >
> > What's more, this change can allow the api to evolve and will not
> > influence the users that call this api.
> >
> > Fixes: 833c95456a70 ("device coredump: add new device coredump class")
> > Signed-off-by: Duoming Zhou <[email protected]>
> > ---
> > Changes in v9:
> > - Add context check in dev_coredumpm().
> >
> > drivers/base/devcoredump.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c
> > index f4d794d6bb8..806ee872f5f 100644
> > --- a/drivers/base/devcoredump.c
> > +++ b/drivers/base/devcoredump.c
> > @@ -255,6 +255,9 @@ void dev_coredumpm(struct device *dev, struct module *owner,
> > struct devcd_entry *devcd;
> > struct device *existing;
> >
> > + if (!gfpflags_normal_context(gfp))
> > + goto free;
>
> Wait, this just broke things for no good reason if a caller happened to
> have set a GFP flag that we do not like at the moment.
>
> What code now does not work that previously did work with this change?
I found that all users in the kernel call the dev_coredumpv(), dev_coredumpm() and
dev_coredumpsg() with "GFP_KERNEL". So this change will not influence the existing users.
Best regards,
Duoming Zhou
On Mon, Sep 26, 2022 at 02:38:28PM +0800, [email protected] wrote:
> Hello,
>
> On Mon, 26 Sep 2022 08:27:58 +0200 Greg KH wrote:
>
> > On Mon, Sep 26, 2022 at 02:16:09PM +0800, Duoming Zhou wrote:
> > > The dev_coredumpm(), dev_coredumpv() and dev_coredumpsg() 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 adds context check in dev_coredumpm() in order to show
> > > dev_coredumpm() and its callers could not be used in atomic context.
> > >
> > > What's more, this change can allow the api to evolve and will not
> > > influence the users that call this api.
> > >
> > > Fixes: 833c95456a70 ("device coredump: add new device coredump class")
> > > Signed-off-by: Duoming Zhou <[email protected]>
> > > ---
> > > Changes in v9:
> > > - Add context check in dev_coredumpm().
> > >
> > > drivers/base/devcoredump.c | 3 +++
> > > 1 file changed, 3 insertions(+)
> > >
> > > diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c
> > > index f4d794d6bb8..806ee872f5f 100644
> > > --- a/drivers/base/devcoredump.c
> > > +++ b/drivers/base/devcoredump.c
> > > @@ -255,6 +255,9 @@ void dev_coredumpm(struct device *dev, struct module *owner,
> > > struct devcd_entry *devcd;
> > > struct device *existing;
> > >
> > > + if (!gfpflags_normal_context(gfp))
> > > + goto free;
> >
> > Wait, this just broke things for no good reason if a caller happened to
> > have set a GFP flag that we do not like at the moment.
> >
> > What code now does not work that previously did work with this change?
>
> I found that all users in the kernel call the dev_coredumpv(), dev_coredumpm() and
> dev_coredumpsg() with "GFP_KERNEL". So this change will not influence the existing users.
Great, so there is no need for this, and it does not "fix" any commit.
confused,
greg k-h