[PATCH v2] cifs: clear PF_MEMALLOC before exiting demultiplex thread

From: Eric Biggers
Date: Sun Mar 08 2020 - 01:17:33 EST


From: Eric Biggers <ebiggers@xxxxxxxxxx>

Leaving PF_MEMALLOC set when exiting a kthread causes it to remain set
during do_exit(). That can confuse things. For example, if BSD process
accounting is enabled, then it's possible for do_exit() to end up
calling ext4_write_inode(). That triggers the
WARN_ON_ONCE(current->flags & PF_MEMALLOC) there, as it assumes
(appropriately) that inodes aren't written when allocating memory.

This case was reported by syzbot at
https://lkml.kernel.org/r/0000000000000e7156059f751d7b@xxxxxxxxxxx

Fix this in cifs_demultiplex_thread() by using the helper functions to
save and restore PF_MEMALLOC.

Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
---

v2: added missing include of <linux/sched/mm.h>
(I missed that I didn't actually have CONFIG_CIFS set...)

fs/cifs/connect.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 4804d1df8c1c..97b8eb585cf9 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -21,6 +21,7 @@
#include <linux/fs.h>
#include <linux/net.h>
#include <linux/string.h>
+#include <linux/sched/mm.h>
#include <linux/sched/signal.h>
#include <linux/list.h>
#include <linux/wait.h>
@@ -1164,8 +1165,9 @@ cifs_demultiplex_thread(void *p)
struct task_struct *task_to_wake = NULL;
struct mid_q_entry *mids[MAX_COMPOUND];
char *bufs[MAX_COMPOUND];
+ unsigned int noreclaim_flag;

- current->flags |= PF_MEMALLOC;
+ noreclaim_flag = memalloc_noreclaim_save();
cifs_dbg(FYI, "Demultiplex PID: %d\n", task_pid_nr(current));

length = atomic_inc_return(&tcpSesAllocCount);
@@ -1320,6 +1322,7 @@ cifs_demultiplex_thread(void *p)
set_current_state(TASK_RUNNING);
}

+ memalloc_noreclaim_restore(noreclaim_flag);
module_put_and_exit(0);
}

--
2.25.1