[Issue]
There is a scenario which efi_pstore may hang up:
- cpuA grabs efivars->lock
- cpuB panics and calls smp_send_stop
- smp_send_stop sends IRQ to cpuA
- after 1 second, cpuB gives up on cpuA and sends an NMI instead
- cpuA is now in an NMI handler while still holding efivars->lock
- cpuB is deadlocked
This case may happen if a firmware has a bug and
cpuA is stuck talking with it.
[Solution]
This patch changes a spin_lock to a spin_trylock in non-blocking paths.
and if the spin_lock has already taken by another cpu,
it returns without accessing to a firmware to avoid the deadlock.
Signed-off-by: Seiji Aguchi <seiji.agu...@hds.com>
---
drivers/firmware/efivars.c | 11 ++++++++++-
fs/pstore/platform.c | 6 +++---
include/linux/pstore.h | 6 ++++++
3 files changed, 19 insertions(+), 4 deletions(-)
diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
index 52c5d89..25d464e 100644
--- a/drivers/firmware/efivars.c
+++ b/drivers/firmware/efivars.c
@@ -1210,7 +1210,16 @@ static int efi_pstore_write(enum pstore_type_id type,
u64 storage_space, remaining_space, max_variable_size;
efi_status_t status = EFI_NOT_FOUND;
- spin_lock(&efivars->lock);
+ if (pstore_is_non_blocking_path(reason)) {
+ /*
+ * If the lock is taken by another cpu in non-blocking path,
+ * this driver returns without entering firmware to avoid
+ * hanging up.
+ */
+ if (!spin_trylock(&efivars->lock))
+ return -EBUSY;
+ } else
+ spin_lock(&efivars->lock);
/*
* Check if there is a space enough to log.
diff --git a/fs/pstore/platform.c b/fs/pstore/platform.c
index b87d1de..f959eba 100644
--- a/fs/pstore/platform.c
+++ b/fs/pstore/platform.c
@@ -96,7 +96,7 @@ static const char *get_reason_str(enum kmsg_dump_reason
reason)
}
}
-static bool is_non_blocking_path(enum kmsg_dump_reason reason)
+bool pstore_is_non_blocking_path(enum kmsg_dump_reason reason)
{
/*
* In case of NMI path, pstore shouldn't be blocked
@@ -134,7 +134,7 @@ static void pstore_dump(struct kmsg_dumper *dumper,
why = get_reason_str(reason);
- if (is_non_blocking_path(reason)) {
+ if (pstore_is_non_blocking_path(reason)) {
is_locked = spin_trylock_irqsave(&psinfo->buf_lock, flags);
if (!is_locked) {
pr_err("pstore dump routine blocked in %s path, may
corrupt error record\n"
@@ -165,7 +165,7 @@ static void pstore_dump(struct kmsg_dumper *dumper,
total += hsize + len;
part++;
}
- if (is_non_blocking_path(reason)) {
+ if (pstore_is_non_blocking_path(reason)) {
if (is_locked)
spin_unlock_irqrestore(&psinfo->buf_lock, flags);
} else
diff --git a/include/linux/pstore.h b/include/linux/pstore.h
index 1788909..595a040 100644
--- a/include/linux/pstore.h
+++ b/include/linux/pstore.h
@@ -68,12 +68,18 @@ struct pstore_info {
#ifdef CONFIG_PSTORE
extern int pstore_register(struct pstore_info *);
+extern bool pstore_is_non_blocking_path(enum kmsg_dump_reason reason);
#else
static inline int
pstore_register(struct pstore_info *psi)
{
return -ENODEV;
}
+static inline bool
+pstore_is_non_blocking_path(enum kmsg_dump_reason reason)
+{
+ return false;
+}
#endif
#endif /*_LINUX_PSTORE_H*/
--
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
