quota: do not leak info to user-space
PATCH 1/1 quota: do not leak info to user-space
From: Cong Wang <[email protected]>
There is a hole in struct fs_quota_stat, so we have to zero the struct on stack before copying it to user-space.
---
fs/quota/quota.c +1 -0
@@ -211,6 +211,7 @@ static int quota_getxstate(struct super_block *sb, void __user *addr)if (!sb->s_qcop->get_xstate)return -ENOSYS ; +memset (&fqs,0 ,sizeof (fqs)); ret = sb->s_qcop->get_xstate (sb, &fqs);if (!ret &©_to_user (addr, &fqs,sizeof (fqs)))return -EFAULT ;
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
There is a hole in struct
fs_quota_stat, so we have to zero the struct on stack before copying it to user-space.Cc Jan Kara <[email protected]>Signed-off-by Cong Wang <[email protected]>
Good point. I've merged the patch.
BTW for XFS folks: The structure definition looks somewhat odd (unaligned definition of qs_flags, qs_uquota starts only at 32-bit boundary although it has 64-bit fields in it) and I wouldn't be surprised if it needed compat wrapper for 32-bit apps on some architectures...
Honza
Quoted 14 lines
---
fs/quota/quota.c +1
-0@@ -211,6 +211,7 @@ static int quota_getxstate(struct super_block *sb, void __user *addr)if (!sb->s_qcop->get_xstate)return -ENOSYS ; +memset (&fqs,0 ,sizeof (fqs)); ret = sb->s_qcop->get_xstate (sb, &fqs);if (!ret &©_to_user (addr, &fqs,sizeof (fqs)))return -EFAULT ;
Signature · 7 lines
Jan Kara <[email protected]> SUSE Labs, CR _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Quoted 9 lines · Jan Kara
Quoted 7 lines · Cong Wang
From: Cong Wang <[email protected]>
There is a hole in struct
fs_quota_stat, so we have to zero the struct on stack before copying it to user-space.Cc Jan Kara <[email protected]>Signed-off-by Cong Wang <[email protected]>Good point. I've merged the patch.
Ah, now I've noticed that XFS (the only user of the callback you are
fixing) is zeroing the structure on its own (xfs_qm_scall_getqstat). So there's no real problem. I'm somewhat wondering whether clearing the field in the place where you did it isn't more future-proof but usually we don't pass in prezeroed buffers so I've decided to leave things as they are.
Honza
Quoted 23 lines
BTW for XFS folks: The structure definition looks somewhat odd (unaligned definition of
qs_flags,qs_uquotastarts only at 32-bit boundary although it has 64-bit fields in it) and I wouldn't be surprised if it needed compat wrapper for 32-bit apps on some architectures...Honza
Quoted 14 lines
---
fs/quota/quota.c +1
-0@@ -211,6 +211,7 @@ static int quota_getxstate(struct super_block *sb, void __user *addr)if (!sb->s_qcop->get_xstate)return -ENOSYS ; +memset (&fqs,0 ,sizeof (fqs)); ret = sb->s_qcop->get_xstate (sb, &fqs);if (!ret &©_to_user (addr, &fqs,sizeof (fqs)))return -EFAULT ;Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR
Signature · 7 lines
Jan Kara <[email protected]> SUSE Labs, CR _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Quoted 7 lines · Jan Kara
Quoted 6 lines · Jan Kara
Quoted 5 lines · Cong Wang
From: Cong Wang <[email protected]>
There is a hole in struct
fs_quota_stat, so we have to zero the struct on stack before copying it to user-space.
Quoted 9 lines
Quoted 3 lines
Quoted 2 lines
Cc Jan Kara <[email protected]>Signed-off-by Cong Wang <[email protected]>Good point. I've merged the patch.
Ah, now I've noticed that XFS (the only user of the callback you are
fixing) is zeroing the structure on its own (xfs_qm_scall_getqstat). So there's no real problem. I'm somewhat wondering whether clearing the field in the place where you did it isn't more future-proof but usually we don't pass in prezeroed buffers so I've decided to leave things as they are.
You are right. Thanks for looking into this!
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs