OOPS in request_key.c bisected (and then refound)
A message that is not in the archive (sent off-list or not yet received): [email protected]
Copying lkml on this additional feedback on the patch to fix the oops introduced in April by the keyctl subsystem
Quoted 93 lines · Rutger Nijlunsing
When trying to mount an CIFS share with SPGENO on Debian, I got nothing. Digging deeping revealing on OOPS at function call_sbin_request_key+0x166/0x255in 2.6.27 which was not there in 2.6.25. Bisecting this with the simplest command generating the OOPS, which was taken from 'man keyctl':
keyctl request2 user debug:yyyy spoon
took about 4 hours on the evolutionary dead-end Pentium 4 and returned commit:
commit
69664cf16af4f31cd54d77948a4baf9c7e0ca7b9Author: David Howells <[email protected]> Date: Tue Apr 29 01:01:31 2008 -0700keys: don't generate user and user session keyrings unless they're accessed
...as the culprit.
Googling for this revealed that this OOPS had been reported before in May, that a patch was written, tested and considered OK:
http://lists.samba.org/archive/linux-cifs-client/2008-May/003001.htmlApplying this patch is still the right thing to do since it made the OOPS disappear. Hopefully this will solve my SPNEGO problems, but that's a second concern. Here is the patch again, together with additional Tested-bys:
---
KEYS: Make request key instantiate the per-user keyringsMake
request_key()instantiate the per-user keyrings so that it doesn't oops if it needs to get hold of the user session keyring because there isn't a session keyring in place.Signed-off-by David Howells <[email protected]>Tested-by Steve French <[email protected]>Tested-by Rutger Nijlunsing <[email protected]>---
security/keys/internal.h | 1 + security/keys/process_keys.c | 2 +- security/keys/request_key.c | 4 ++++ 3 files changed, 6 insertions(+), 1 deletions(-)security/keys/internal.h +1
-0@@ -108,6 +108,7 @@ extern key_ref_t search_process_keyrings(struct key_type *type,extern struct key *find_keyring_by_name (const char *name,bool skip_perm_check); +extern int install_user_keyrings (struct task_struct *tsk);extern int install_thread_keyring (struct task_struct *tsk);extern int install_process_keyring (struct task_struct *tsk);security/keys/process_keys.c +1
-1@@ -40,7 +40,7 @@ struct key_user root_key_user = {/* * install user and user session keyrings for a particular UID */ -static int install_user_keyrings (struct task_struct *tsk) +int install_user_keyrings (struct task_struct *tsk) {struct user_struct *user = tsk->user;struct key *uid_keyring, *session_keyring;security/keys/request_key.c +4
-1@@ -74,6 +74,10 @@ static int call_sbin_request_key(struct key_construction *cons,kenter ("{%d},{%d},%s" , key->serial, authkey->serial, op); + ret =install_user_keyrings (tsk); +if (ret <0 ) + goto error_alloc; +/* allocate a new session keyring */ sprintf (desc,"_req.%u" , key->serial); --Rutger Nijlunsing ---------------------------------- eludias ed dse.nl never attribute to a conspiracy which can be explained by incompetence
----------------------------------------------------------------------
Signature · 3 lines
Thanks, Steve
This patch was confirmed by a user here to fix the oops (happened while using mount.cifs with kerberos).
Is there a reason not to merge it (and into stable too) ?
Quoted 97 lines · Steve French
Copying lkml on this additional feedback on the patch to fix the oops introduced in April by the keyctl subsystem
Quoted 92 lines · Rutger Nijlunsing
When trying to mount an CIFS share with SPGENO on Debian, I got nothing. Digging deeping revealing on OOPS at function call_sbin_request_key+0x166/0x255in 2.6.27 which was not there in 2.6.25. Bisecting this with the simplest command generating the OOPS, which was taken from 'man keyctl':
keyctl request2 user debug:yyyy spoon
took about 4 hours on the evolutionary dead-end Pentium 4 and returned commit:
commit
69664cf16af4f31cd54d77948a4baf9c7e0ca7b9Author: David Howells <[email protected]> Date: Tue Apr 29 01:01:31 2008 -0700keys: don't generate user and user session keyrings unless they're accessed
...as the culprit.
Googling for this revealed that this OOPS had been reported before in May, that a patch was written, tested and considered OK:
http://lists.samba.org/archive/linux-cifs-client/2008-May/003001.htmlApplying this patch is still the right thing to do since it made the OOPS disappear. Hopefully this will solve my SPNEGO problems, but that's a second concern. Here is the patch again, together with additional Tested-bys:
---
KEYS: Make request key instantiate the per-user keyringsMake
request_key()instantiate the per-user keyrings so that it doesn't oops if it needs to get hold of the user session keyring because there isn't a session keyring in place.Signed-off-by David Howells <[email protected]>Tested-by Steve French <[email protected]>Tested-by Rutger Nijlunsing <[email protected]>---
security/keys/internal.h | 1 + security/keys/process_keys.c | 2 +- security/keys/request_key.c | 4 ++++ 3 files changed, 6 insertions(+), 1 deletions(-)security/keys/internal.h +1
-0@@ -108,6 +108,7 @@ extern key_ref_t search_process_keyrings(struct key_type *type,extern struct key *find_keyring_by_name (const char *name,bool skip_perm_check); +extern int install_user_keyrings (struct task_struct *tsk);extern int install_thread_keyring (struct task_struct *tsk);extern int install_process_keyring (struct task_struct *tsk);security/keys/process_keys.c +1
-1@@ -40,7 +40,7 @@ struct key_user root_key_user = {/* * install user and user session keyrings for a particular UID */ -static int install_user_keyrings (struct task_struct *tsk) +int install_user_keyrings (struct task_struct *tsk) {struct user_struct *user = tsk->user;struct key *uid_keyring, *session_keyring;security/keys/request_key.c +4
-1@@ -74,6 +74,10 @@ static int call_sbin_request_key(struct key_construction *cons,kenter ("{%d},{%d},%s" , key->serial, authkey->serial, op); + ret =install_user_keyrings (tsk); +if (ret <0 ) + goto error_alloc; +/* allocate a new session keyring */ sprintf (desc,"_req.%u" , key->serial); --Rutger Nijlunsing ---------------------------------- eludias ed dse.nl never attribute to a conspiracy which can be explained by incompetence
----------------------------------------------------------------------
Quoted 3 lines · Pascal Terjan
This patch was confirmed by a user here to fix the oops (happened while using mount.cifs with kerberos).
Is there a reason not to merge it (and into stable too) ?
I thought it was upstream, but I think I must've been looking at the wrong GIT tree. I'll push it now.
David