DLM regression in 64-bit 3.7.x Kernel?
Hi,
I have recently upgraded my development cluster from 3.6.x to 3.7.1 kernel and clvmd stopped working (all locking operation result in 'Invalid argument'). I have traced the problem to this call:
write(8, "\6\0\0\0\0\0\0\0\1\0\0\0\1\1\0\0\5@\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\320\307\351\326\277\177\0\0\20\17\36\331\277\177\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\313\351\326\277\177\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0Hlk5NeaVF0qhDF20RBq61EZaIj5yyUJgWxxs4ntZedP34Rj3E2wadZfVwtOcvpKb", 184) = -1 EINVAL (Invalid argument)
Kernel sources show only a few conditions which can result in the EINVAL error and the one introduced by the 'dlm: check the maximum size of a request from user' patch seemed suspicious to me.
Should a 64-bit kernel really refuse requests using the 64-bit dlm_write_request struct when compiled with CONFIG_COMPAT on?
To verify my suspicions I have compiled the 3.7.9 with a little patch dropping the '(count > sizeof(struct dlm_write_request32) + DLM_RESNAME_MAXLEN)' check, which solved the problem for me.
The patch will follow, though I am still not sure this is the solution.
Greets,
Jacek
PATCH 1/1 dlm: Allow 64-bit request when CONFIG_COMPAT is on
Even if kernel is compiled for 32-bit ABI compatibility it still should able to process 64-bit dlm_write_request struct.
The regression was introduced by:
b75bc91 dlm: check the maximum size of a request from user
--- fs/dlm/user.c | 4 ---- 1 file changed, 4 deletions(-)
fs/dlm/user.c +0 -4
@@ -503,11 +503,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,#endif return -EINVAL ; -#ifdef CONFIG_COMPAT -if (count >sizeof (struct dlm_write_request32 ) +DLM_RESNAME_MAXLEN ) -#elseif (count >sizeof (struct dlm_write_request ) +DLM_RESNAME_MAXLEN ) -#endif return -EINVAL ; kbuf =kzalloc (count +1 ,GFP_NOFS );
Signature · 1 line
1.8.0.1
David Teigland
maintainerQuoted 14 lines · Jacek Konieczny
Hi,
I have recently upgraded my development cluster from 3.6.x to 3.7.1 kernel and clvmd stopped working (all locking operation result in 'Invalid argument'). I have traced the problem to this call:
write(8, "\6\0\0\0\0\0\0\0\1\0\0\0\1\1\0\0\5@\0\0\1\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\320\307\351\326\277\177\0\0\20\17\36\331\277\177\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\313\351\326\277\177\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0Hlk5NeaVF0qhDF20RBq61EZaIj5yyUJgWxxs4ntZedP34Rj3E2wadZfVwtOcvpKb", 184) = -1 EINVAL (Invalid argument)
Kernel sources show only a few conditions which can result in the EINVAL error and the one introduced by the 'dlm: check the maximum size of a request from user' patch seemed suspicious to me.
Quoted 8 lines
Should a 64-bit kernel really refuse requests using the 64-bit
dlm_write_requeststruct when compiled withCONFIG_COMPATon?To verify my suspicions I have compiled the 3.7.9 with a little patch dropping the '(count >
sizeof(struct dlm_write_request32)+DLM_RESNAME_MAXLEN)' check, which solved the problem for me.The patch will follow, though I am still not sure this is the solution.
Hi, the fix is already upstream: d4b0bcf32b946590afd29e202d6a399b84fe6c67 Dave