PATCH: ALSA: hda: fix possible null dereference
Upgrading on Gentoo hardened from 4.0.8 to 4.1.6 (with intermediate 4.1.4), I noticed a NULL pointer derefence during booting, and tackled it down to
static int add_std_chmaps(struct hda_codec *codec)
in file sound/pci/hda/hda_codec.c
With commits bbbc7e85 and 751e2216, a for-loop was restated using list_for_each_entry(...); originally, a local pcm* was extracted from an array and checked against NULL, if not-NULL then passed along to snd_pcm_add_chmap_ctls(...).
Now, pcm->pcm is passed to snd_pcm_add_chmap_ctls, but the NULL pointer check still uses the “upper level” pcm, not the “in-loop” pcm->pcm.
Please have a look at the attached (rather small) suggestion to fix this problem.
For a history of the function in question, have a look at lines 2430ff for
git diff v4.0 v4.1 sound/pci/hda/hda_codec.c
The two commits that touched this function are
Commit bbbc7e8502c9 ("ALSA: hda - Allocate hda_pcm objects dynamically") Commit 751e2216899c ("ALSA: hda: fix possible null dereference")
where the latter only fixes the order of two fixes, but not the pcm / pcm->pcm confusion.
Using v4.1-descendant sources, boot got OOPS like this (two bug-hunting printk's added up front):
Quoted 6 lines
pcm:
ffff88040cf83400pcm->pcm: (nil)BUG: unable to handle kernel NULL pointer dereference at 0000000000000018 IP: [<ffffffffa07feab9>] snd_pcm_add_chmap_ctls+0x119/0x1f0 [snd_pcm] PGD 0 Oops: 0000 [#1] PREEMPT SMP
NULL checking pcm->pcm instead of pcm fixes this problem.
Greetings and thanks for your time, Markus
---
pci/hda/hda_codec.c.orig renamed from pci/hda/hda_codec.c +1 -1
@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_cod struct snd_pcm_chmap *chmap; const struct snd_pcm_chmap_elem *elem; - if (!pcm->pcm || pcm->own_chmap || + if (!pcm || pcm->own_chmap || !hinfo->substreams) continue; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
Quoted 39 lines · Markus Osterhoff
Upgrading on Gentoo hardened from 4.0.8 to 4.1.6 (with intermediate 4.1.4), I noticed a NULL pointer derefence during booting, and tackled it down to
static intadd_std_chmaps(struct hda_codec *codec)
in filesound/pci/hda/hda_codec.cWith commits
bbbc7e85and751e2216, a for-loop was restated usinglist_for_each_entry(...); originally, a local pcm* was extracted from an array and checked against NULL, if not-NULL then passed along tosnd_pcm_add_chmap_ctls(...).Now,
pcm->pcmis passed tosnd_pcm_add_chmap_ctls, but the NULL pointer check still uses the “upper level” pcm, not the “in-loop”pcm->pcm.Please have a look at the attached (rather small) suggestion to fix this problem.
For a history of the function in question, have a look at lines 2430ff for
git diff v4.0 v4.1sound/pci/hda/hda_codec.cThe two commits that touched this function are
Commitbbbc7e8502c9("ALSA: hda - Allocatehda_pcmobjects dynamically") Commit751e2216899c("ALSA: hda: fix possible null dereference")where the latter only fixes the order of two fixes, but not the pcm /
pcm->pcmconfusion.Using v4.1-descendant sources, boot got OOPS like this (two bug-hunting printk's added up front):
Quoted 6 lines
pcm:
ffff88040cf83400pcm->pcm: (nil)BUG: unable to handle kernel NULL pointer dereference at 0000000000000018 IP: [<ffffffffa07feab9>] snd_pcm_add_chmap_ctls+0x119/0x1f0 [snd_pcm] PGD 0 Oops: 0000 [#1] PREEMPT SMP
Quoted 6 lines
NULL checking
pcm->pcminstead of pcm fixes this problem.Greetings and thanks for your time, Markus
Signed-off-by Markus Osterhoff <[email protected]>
Thanks for the patch. But this was already fixed by commit 751e2216899c in 4.1-rc1.
Takashi
Quoted 6 lines
---
pci/hda/hda_codec.c.orig static int add_std_chmaps(struct hda_cod · line 3178
Quoted 8 lines
pci/hda/hda_codec.c.orig renamed from pci/hda/hda_codec.c +1
-1@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_cod struct snd_pcm_chmap *chmap; const struct snd_pcm_chmap_elem *elem; - if (!pcm->pcm || pcm->own_chmap || + if (!pcm || pcm->own_chmap || !hinfo->substreams) continue; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;
_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
Dear Takashi,
maybe I am mistaken or have not made my patch clear, please let me apologise.
Quoted 10 lines · Takashi Iwai
Quoted 8 lines · Markus Osterhoff
With commits
bbbc7e85and751e2216, a for-loop was restated usinglist_for_each_entry(...); originally, a local pcm* was extracted from an array and checked against NULL, if not-NULL then passed along tosnd_pcm_add_chmap_ctls(...).For a history of the function in question, have a look at lines 2430ff for
git diff v4.0 v4.1sound/pci/hda/hda_codec.c
Quoted 8 lines
Quoted 5 lines
The two commits that touched this function are
Commitbbbc7e8502c9("ALSA: hda - Allocatehda_pcmobjects dynamically") Commit751e2216899c("ALSA: hda: fix possible null dereference")where the latter only fixes the order of two fixes, but not the pcm /
pcm->pcmconfusion.Thanks for the patch. But this was already fixed by commit
751e2216899cin 4.1-rc1.
So I re-checked and diff'ed against the current v4.2-rc8 and found that the if-statement reads, /sound/pci/hda/hda_codec.c, line 3172:
if (!pcm || pcm->own_chmap || while I propose if (!pcm->pcm || pcm->own_chmap ||
With the mentioned 751e2216899c I do get a NULL dereference (for whatever reason), because pcm->pcm can indeed by NULL inside the loop. Please reconsider the patch:
Quoted 1 line
Takashi
Thanks for your time, Markus
--- sound/pci/hda/hda_codec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
sound/pci/hda/hda_codec.c +1 -1
@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_codec *codec)struct snd_pcm_chmap *chmap;const struct snd_pcm_chmap_elem *elem; -if (!pcm || pcm->own_chmap || +if (!pcm->pcm || pcm->own_chmap || !hinfo->substreams)continue ; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;
Quoted 23 lines · Markus Osterhoff
Dear Takashi,
maybe I am mistaken or have not made my patch clear, please let me apologise.
Quoted 15 lines · Takashi Iwai
Quoted 13 lines · Markus Osterhoff
With commits
bbbc7e85and751e2216, a for-loop was restated usinglist_for_each_entry(...); originally, a local pcm* was extracted from an array and checked against NULL, if not-NULL then passed along tosnd_pcm_add_chmap_ctls(...).For a history of the function in question, have a look at lines 2430ff for
git diff v4.0 v4.1sound/pci/hda/hda_codec.cThe two commits that touched this function are
Commitbbbc7e8502c9("ALSA: hda - Allocatehda_pcmobjects dynamically") Commit751e2216899c("ALSA: hda: fix possible null dereference")where the latter only fixes the order of two fixes, but not the pcm /
pcm->pcmconfusion.
Quoted 7 lines
Quoted 2 lines
Thanks for the patch. But this was already fixed by commit
751e2216899cin 4.1-rc1.So I re-checked and diff'ed against the current v4.2-rc8 and found that the if-statement reads, /sound/pci/hda/hda_codec.c, line 3172:
if (!pcm || pcm->own_chmap || while I propose if (!pcm->pcm || pcm->own_chmap ||
Then your previous patch did exactly opposite :)
Quoted 3 lines
With the mentioned
751e2216899cI do get a NULL dereference (for whatever reason), becausepcm->pcmcan indeed by NULL inside the loop. Please reconsider the patch:
Actually you're trying to shoot two things in a shot, and it's confusing. What we need is to *add* the NULL check of pcm->pcm. It's basically different from replacing pcm NULL check itself. The pcm NULL check is superfluous, indeed, and can be removed. But it can be done in a different patch, or mention properly in the changelog.
Could you revisit your patch with the proper description?
thanks,
Takashi
Quoted 24 lines
Quoted 2 lines
Takashi
Thanks for your time, Markus
Signed-off-by Markus Osterhoff <[email protected]>--- sound/pci/hda/hda_codec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)sound/pci/hda/hda_codec.c +1
-1@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_codec *codec)struct snd_pcm_chmap *chmap;const struct snd_pcm_chmap_elem *elem; -if (!pcm || pcm->own_chmap || +if (!pcm->pcm || pcm->own_chmap || !hinfo->substreams)continue ; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;
PATCH RESEND v2 1/1 ALSA: hda: fix possible NULL dereference
After a for-loop was replaced by list_for_each_entry, see Commit bbbc7e8502c9 ("ALSA: hda - Allocate hda_pcm objects dynamically"), Commit 751e2216899c ("ALSA: hda: fix possible null dereference"), a possible NULL pointer dereference has been introduced; this patch adds the NULL check on pcm->pcm, while leaving a potentially superfluous check on pcm itself untouched.
--- sound/pci/hda/hda_codec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
sound/pci/hda/hda_codec.c +1 -1
@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_codec *codec)struct snd_pcm_chmap *chmap;const struct snd_pcm_chmap_elem *elem; -if (!pcm || pcm->own_chmap || +if (!pcm || !pcm->pcm || pcm->own_chmap || !hinfo->substreams)continue ; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;
Re: [PATCH RESEND V2] ALSA: hda: fix possible NULL dereference
Quoted 9 lines · Markus Osterhoff
After a for-loop was replaced by
list_for_each_entry, see Commitbbbc7e8502c9("ALSA: hda - Allocatehda_pcmobjects dynamically"), Commit751e2216899c("ALSA: hda: fix possible null dereference"), a possible NULL pointer dereference has been introduced; this patch adds the NULL check onpcm->pcm, while leaving a potentially superfluous check on pcm itself untouched.Signed-off-by Markus Osterhoff <[email protected]>
The patch couldn't be applied cleanly because the tabs are replaced wrongly with spaces, so I applied it manually now. Please fix your MUA or the way to create a patch at the next time.
thanks,
Takashi
Quoted 18 lines
--- sound/pci/hda/hda_codec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)sound/pci/hda/hda_codec.c +1
-1@@ -3172,7 +3172,7 @@ static int add_std_chmaps(struct hda_codec *codec)struct snd_pcm_chmap *chmap;const struct snd_pcm_chmap_elem *elem; -if (!pcm || pcm->own_chmap || +if (!pcm || !pcm->pcm || pcm->own_chmap || !hinfo->substreams)continue ; elem = hinfo->chmap ? hinfo->chmap : snd_pcm_std_chmaps;