[Dle-develop] [RFC][PATCH v4 -next 4/4] Skip spin_lock of efi_pstore_write() in panic case

[Seiji A. <sei...@hd...>]

This patch skips spin_lock of efi_pstore_write() in panic path for these reasons.
 - efi_pstore_write() is serialized via smp_send_stop().
 - SetVariable runtime service is guaranteed to work correctly if you call
   it a second time when you have taken an exception in trying
   to execute them already.

We can avoid deadlock of spin_lock(&efivars->lock) in panic path and get panic log
reliably.


 Signed-off-by: Seiji Aguchi <sei...@hd...>

---
 drivers/firmware/efivars.c |   13 +++++++++++--
 1 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
index d25599f..657e6f1 100644
--- a/drivers/firmware/efivars.c
+++ b/drivers/firmware/efivars.c
@@ -510,7 +510,15 @@ static int efi_pstore_write(enum pstore_type_id type,
 	sprintf(stub_name, "dump-type%u-%u-", type, part);
 	sprintf(name, "%s%lu", stub_name, get_seconds());
 
-	spin_lock(&efivars->lock);
+	/*
+	 * We don't need to take any locks in panic path for these reasons.
+	 *  - efi_pstore_write() is serialized via smp_send_stop().
+	 *  - SetVariable runtime service is guaranteed to work correctly
+	 *    if you call it a second time when you have taken an exception
+	 *    in trying to execute them already.
+	 */
+	if (reason != KMSG_DUMP_PANIC)
+		spin_lock(&efivars->lock);
 
 	for (i = 0; i < DUMP_NAME_LEN; i++)
 		efi_name[i] = stub_name[i];
@@ -548,7 +556,8 @@ static int efi_pstore_write(enum pstore_type_id type,
 	efivars->ops->set_variable(efi_name, &vendor, PSTORE_EFI_ATTRIBUTES,
 				   size, psi->buf);
 
-	spin_unlock(&efivars->lock);
+	if (reason != KMSG_DUMP_PANIC)
+		spin_unlock(&efivars->lock);
 
 	if (found)
 		efivar_unregister(found);
-- 
1.7.1