On Thu, 07 May 2020 12:13:10 +0200, Greg Kroah-Hartman wrote: > > On Thu, May 07, 2020 at 11:56:22AM +0200, Takashi Iwai wrote: > > On Thu, 07 May 2020 10:23:02 +0200, > > Greg Kroah-Hartman wrote: > > > > > > On Thu, May 07, 2020 at 04:04:25PM +0800, butt3rflyh4ck wrote: > > > > I report a bug (in linux-5.7-rc1) found by syzkaller. > > > > > > > > kernel config: https://github.com/butterflyhack/syzkaller-fuzz/blob/master/v5.7.0-rc1.config > > > > reproducer: https://github.com/butterflyhack/syzkaller-fuzz/blob/master/repro.cprog > > > > > > > > I test the reproducer in linux-5.7-rc4 and crash too. > > > > > > Great, care to create a fix for this and send it to the proper > > > maintainers? That's the best way to get it fixed, otherwise it just > > > goes in the file with the rest of the syzbot reports we are burried > > > under. > > > > Don't worry, I already prepared a fix patch below :) > > > > > > thanks, > > > > Takashi > > > > -- 8< -- > > From: Takashi Iwai <tiwai@xxxxxxx> > > Subject: [PATCH] ALSA: rawmidi: Fix racy buffer resize under concurrent > > accesses > > > > The rawmidi core allows user to resize the runtime buffer via ioctl, > > and this may lead to UAF when performed during concurrent reads or > > writes. > > > > This patch fixes the race by introducing a reference counter for the > > runtime buffer access and returns -EBUSY error when the resize is > > performed concurrently. > > > > Reported-by: butt3rflyh4ck <butterflyhuangxx@xxxxxxxxx> > > Cc: <stable@xxxxxxxxxxxxxxx> > > Link: https://lore.kernel.org/r/CAFcO6XMWpUVK_yzzCpp8_XP7+=oUpQvuBeCbMffEDkpe8jWrfg@xxxxxxxxxxxxxx > > Signed-off-by: Takashi Iwai <tiwai@xxxxxxx> > > --- > > include/sound/rawmidi.h | 1 + > > sound/core/rawmidi.c | 29 ++++++++++++++++++++++++++++- > > 2 files changed, 29 insertions(+), 1 deletion(-) > > > > diff --git a/include/sound/rawmidi.h b/include/sound/rawmidi.h > > index a36b7227a15a..334842daa904 100644 > > --- a/include/sound/rawmidi.h > > +++ b/include/sound/rawmidi.h > > @@ -61,6 +61,7 @@ struct snd_rawmidi_runtime { > > size_t avail_min; /* min avail for wakeup */ > > size_t avail; /* max used buffer for wakeup */ > > size_t xruns; /* over/underruns counter */ > > + int buffer_ref; /* buffer reference count */ > > /* misc */ > > spinlock_t lock; > > wait_queue_head_t sleep; > > diff --git a/sound/core/rawmidi.c b/sound/core/rawmidi.c > > index 20dd08e1f675..4185d9e81e3c 100644 > > --- a/sound/core/rawmidi.c > > +++ b/sound/core/rawmidi.c > > @@ -120,6 +120,17 @@ static void snd_rawmidi_input_event_work(struct work_struct *work) > > runtime->event(runtime->substream); > > } > > > > +/* buffer refcount management: call with runtime->lock held */ > > +static inline void snd_rawmidi_buffer_ref(struct snd_rawmidi_runtime *runtime) > > +{ > > + runtime->buffer_ref++; > > +} > > + > > +static inline void snd_rawmidi_buffer_unref(struct snd_rawmidi_runtime *runtime) > > +{ > > + runtime->buffer_ref--; > > +} > > Why not use the reference count structure? The context accessing the buffer is always with the spinlock, so we don't need expensive atomic ops there. Usually this kind of check can be a simple boolean flag, but in this case, there is one place that goes out of lock due to copy_from/to_user, so a refcount is used in this patch instead. thanks, Takashi
