2011-06-28 21:19:24

by Arnaud Giersch

Subject: [PATCH][Resend] coredump: call vfs_getattr() to get inode attributes

From: Arnaud Giersch <[email protected]>

In do_coredump(), call vfs_getattr() to get inode attributes, and do not
get them directly from the fields of the inode struct.

Without this patch, when dumping core on an NFSv4 mount, and the i_uid
field is not correctly filled at open time, the uid check fails, and an
empty core dump is produced.

This apparently only happens when there was no "core" file before the
dump. If a "core" file owned by the current user is already present, it
is correctly filled.

The reason is that decode_attr_owner() in fs/nfs/nfs4xdr.c is not
allowed to call the idmapper when it receives may_sleep = 0 (see commit
80e52aced138bb41b045a8595a87510f27d8d8c5, and some explanations in
http://article.gmane.org/gmane.linux.nfs/33391).

Signed-off-by: Arnaud Giersch <[email protected]>
---
fs/exec.c | 12 +++++++-----
1 files changed, 7 insertions(+), 5 deletions(-)

diff --git a/fs/exec.c b/fs/exec.c
index 6075a1e..a016756 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -2147,7 +2147,7 @@ void do_coredump(long signr, int exit_code, struct pt_regs *regs)
goto close_fail;
}
} else {
- struct inode *inode;
+ struct kstat stat;

if (cprm.limit < binfmt->min_coredump)
goto fail_unlock;
@@ -2158,8 +2158,10 @@ void do_coredump(long signr, int exit_code, struct pt_regs *regs)
if (IS_ERR(cprm.file))
goto fail_unlock;

- inode = cprm.file->f_path.dentry->d_inode;
- if (inode->i_nlink > 1)
+ if (vfs_getattr(cprm.file->f_path.mnt, cprm.file->f_path.dentry,
+ &stat))
+ goto close_fail;
+ if (stat.nlink > 1)
goto close_fail;
if (d_unhashed(cprm.file->f_path.dentry))
goto close_fail;
@@ -2167,13 +2169,13 @@ void do_coredump(long signr, int exit_code, struct pt_regs *regs)
* AK: actually i see no reason to not allow this for named
* pipes etc, but keep the previous behaviour for now.
*/
- if (!S_ISREG(inode->i_mode))
+ if (!S_ISREG(stat.mode))
goto close_fail;
/*
* Dont allow local users get cute and trick others to coredump
* into their pre-created files.
*/
- if (inode->i_uid != current_fsuid())
+ if (stat.uid != current_fsuid())
goto close_fail;
if (!cprm.file->f_op || !cprm.file->f_op->write)
goto close_fail;

2011-06-28 23:26:24

by Myklebust, Trond

Subject: Re: [PATCH][Resend] coredump: call vfs_getattr() to get inode attributes

On Tue, 2011-06-28 at 23:19 +0200, Arnaud Giersch wrote:
> From: Arnaud Giersch <[email protected]>
>
> In do_coredump(), call vfs_getattr() to get inode attributes, and do not
> get them directly from the fields of the inode struct.
>
> Without this patch, when dumping core on an NFSv4 mount, and the i_uid
> field is not correctly filled at open time, the uid check fails, and an
> empty core dump is produced.
>
> This apparently only happens when there was no "core" file before the
> dump. If a "core" file owned by the current user is already present, it
> is correctly filled.
>
> The reason is that decode_attr_owner() in fs/nfs/nfs4xdr.c is not
> allowed to call the idmapper when it receives may_sleep = 0 (see commit
> 80e52aced138bb41b045a8595a87510f27d8d8c5, and some explanations in
> http://article.gmane.org/gmane.linux.nfs/33391).
>
> Signed-off-by: Arnaud Giersch <[email protected]>

This wants to be fixed in the NFS layer, not the VFS.

What we should do is the following:

1. Save the string versions of user@domain/group@domain in the
struct nfs4_opendata so that we can resolve them from the
process context in _nfs4_proc_open().
2. Fix nfs4_atomic_open() so that it revalidates any inode that has
the NFS_INO_INVALID_ATTR flag set.

Cheers
Trond
--
Trond Myklebust
Linux NFS client maintainer

NetApp
[email protected]
http://www.netapp.com