Hi
I would like to discuss a way to avoid deadlocking of pstore_dump(), kmsg_dump() and platform drivers.
At first, I post pstore_dump() part.
This patch is based on following comments on my previous patch.
http://marc.info/?l=linux-kernel&m=131914554220738&w=2
http://marc.info/?l=linux-kernel&m=132370565114963&w=2
The points are skipping spin_lock in panic case and adding WARN_ON in preparation for future change.
Any comments are welcome.
Seiji
[Patch Description]
This patch does followings.
- Skip a spin_lock of pstore_dump() in panic case because pstore_dump() is serialized via smp_send_stop.
- Add WARN_ON() in "in_nmi() and !panic" case into pstore_dump(). Currently, this condition never
becomes true. But if someone adds new kmsg_dump() into NMI path in the future, error record may corrupt.
We can trap it and complain with it.
Signed-off-by: Seiji Aguchi <[email protected]>
---
fs/pstore/platform.c | 26 ++++++++++++++++----------
1 files changed, 16 insertions(+), 10 deletions(-)
diff --git a/fs/pstore/platform.c b/fs/pstore/platform.c index 82c585f..e5a0354 100644
--- a/fs/pstore/platform.c
+++ b/fs/pstore/platform.c
@@ -31,6 +31,8 @@
#include <linux/hardirq.h>
#include <linux/workqueue.h>
+#include <asm-generic/bug.h>
+
#include "internal.h"
/*
@@ -107,15 +109,21 @@ static void pstore_dump(struct kmsg_dumper *dumper,
int hsize, ret;
unsigned int part = 1;
unsigned long flags = 0;
- int is_locked = 0;
why = get_reason_str(reason);
- if (in_nmi()) {
- is_locked = spin_trylock(&psinfo->buf_lock);
- if (!is_locked)
- pr_err("pstore dump routine blocked in NMI, may corrupt error record\n");
- } else
+ /*
+ * Currently, this condition never becomes true.
+ * But if someone adds new kmsg_dump() into NMI path in the future,
+ * error record may corrupt. We can trap it and complain with it.
+ */
+ WARN_ON(in_nmi() && reason != KMSG_DUMP_PANIC);
+
+ /*
+ * pstore_dump() is serialized in panic case.
+ * So, we don't need to take any locks.
+ */
+ if (reason != KMSG_DUMP_PANIC)
spin_lock_irqsave(&psinfo->buf_lock, flags);
oopscount++;
while (total < kmsg_bytes) {
@@ -145,10 +153,8 @@ static void pstore_dump(struct kmsg_dumper *dumper,
total += l1_cpy + l2_cpy;
part++;
}
- if (in_nmi()) {
- if (is_locked)
- spin_unlock(&psinfo->buf_lock);
- } else
+
+ if (reason != KMSG_DUMP_PANIC)
spin_unlock_irqrestore(&psinfo->buf_lock, flags); }
--
1.7.1