[1/1] coredump: Fix null pointer dereference when kernel.core_pattern is "|"

Commit Message

Matthew Ruffell Feb. 20, 2020, 5:10 a.m. UTC

Since 315c692, a null pointer dereference can be triggered when the
kernel.core_pattern string is set to "|", and a user executes a program which
crashes.

This is caused by a subtle change in parameters sent to
call_usermodehelper_exec(), as sub_info->path will be set to cn.corename, which
has not changed from its initial value of '\0'. call_usermodehelper_exec()
will return 0 upon strlen() finding that sub_info->path has zero length.

The fix is to add a length check for cn.corename when we check the return code
from call_usermodehelper_exec(). This restores the expected semantics as seen
before 315c692, with the message "Core dump to | pipe failed" output to dmesg
and the coredump being aborted.

Fixes: 315c692 ("coredump: split pipe command whitespace before expanding template")
Signed-off-by: Matthew Ruffell <[e-mail]>
---
 fs/coredump.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Patch

diff --git a/fs/coredump.c b/fs/coredump.c
index b1ea7dfbd149..ca5976e81d8a 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -686,7 +686,7 @@  void do_coredump(const kernel_siginfo_t *siginfo)
 							  UMH_WAIT_EXEC);
 
 		kfree(helper_argv);
-		if (retval) {
+		if (retval || strlen(cn.corename) == 0) {
 			printk(KERN_INFO "Core dump to |%s pipe failed\n",
 			       cn.corename);
 			goto close_fail;