[MTD] [NOR] fix cfi_cmdset_0001 FL_SYNCING race (take 2)
The patch fixes CFI issue with multipartitional devices leading to the
set of errors or even deadlock. The problem is CFI FL_SYNCING state race
with flash operations (e.g. erase suspend). It is reproduced by running
intensive writes on one JFFS2 partition and simultaneously performing
mount/unmount cycle on another partition of the same chip.
Signed-off-by: Alexander Belyakov <abelyako@googlemail.com>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
diff --git a/drivers/mtd/chips/cfi_cmdset_0001.c b/drivers/mtd/chips/cfi_cmdset_0001.c
index 5157e3c..c93a8be 100644
--- a/drivers/mtd/chips/cfi_cmdset_0001.c
+++ b/drivers/mtd/chips/cfi_cmdset_0001.c
@@ -725,6 +725,10 @@
struct cfi_pri_intelext *cfip = cfi->cmdset_priv;
unsigned long timeo = jiffies + HZ;
+ /* Prevent setting state FL_SYNCING for chip in suspended state. */
+ if (mode == FL_SYNCING && chip->oldstate != FL_READY)
+ goto sleep;
+
switch (chip->state) {
case FL_STATUS:
@@ -830,8 +834,9 @@
DECLARE_WAITQUEUE(wait, current);
retry:
- if (chip->priv && (mode == FL_WRITING || mode == FL_ERASING
- || mode == FL_OTP_WRITE || mode == FL_SHUTDOWN)) {
+ if (chip->priv &&
+ (mode == FL_WRITING || mode == FL_ERASING || mode == FL_OTP_WRITE
+ || mode == FL_SHUTDOWN) && chip->state != FL_SYNCING) {
/*
* OK. We have possibility for contention on the write/erase
* operations which are global to the real chip and not per
@@ -881,6 +886,14 @@
return ret;
}
spin_lock(&shared->lock);
+
+ /* We should not own chip if it is already
+ * in FL_SYNCING state. Put contender and retry. */
+ if (chip->state == FL_SYNCING) {
+ put_chip(map, contender, contender->start);
+ spin_unlock(contender->mutex);
+ goto retry;
+ }
spin_unlock(contender->mutex);
}