summaryrefslogtreecommitdiff
path: root/drivers/mtd/ubi/wl.c
diff options
context:
space:
mode:
authorArtem Bityutskiy <Artem.Bityutskiy@nokia.com>2009-05-24 15:13:34 +0400
committerArtem Bityutskiy <Artem.Bityutskiy@nokia.com>2009-06-02 14:53:35 +0400
commitb86a2c56e512f46d140a4bcb4e35e8a7d4a99a4b (patch)
tree59c3e036dfd767b73e700bd7fd8cb4bee15c4f58 /drivers/mtd/ubi/wl.c
parent87960c0b12d0c5a0b37e0c79aef77aa1a0b10d44 (diff)
downloadlinux-b86a2c56e512f46d140a4bcb4e35e8a7d4a99a4b.tar.xz
UBI: do not switch to R/O mode on read errors
This patch improves UBI errors handling. ATM UBI switches to R/O mode when the WL worker fails to read the source PEB. This means that the upper layers (e.g., UBIFS) has no chances to unmap the erroneous PEB and fix the error. This patch changes this behaviour and makes UBI put PEBs like this into a separate RB-tree, thus preventing the WL worker from hitting the same read errors again and again. But there is a 10% limit on a maximum amount of PEBs like this. If there are too much of them, UBI switches to R/O mode. Additionally, this patch teaches UBI not to panic and switch to R/O mode if after a PEB has been copied, the target LEB cannot be read back. Instead, now UBI cancels the operation and schedules the target PEB for torturing. The error paths has been tested by ingecting errors into 'ubi_eba_copy_leb()'. Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Diffstat (limited to 'drivers/mtd/ubi/wl.c')
-rw-r--r--drivers/mtd/ubi/wl.c45
1 files changed, 38 insertions, 7 deletions
diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 793882ba2a6e..9d1d3595a240 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -55,8 +55,8 @@
*
* As it was said, for the UBI sub-system all physical eraseblocks are either
* "free" or "used". Free eraseblock are kept in the @wl->free RB-tree, while
- * used eraseblocks are kept in @wl->used or @wl->scrub RB-trees, or
- * (temporarily) in the @wl->pq queue.
+ * used eraseblocks are kept in @wl->used, @wl->erroneous, or @wl->scrub
+ * RB-trees, as well as (temporarily) in the @wl->pq queue.
*
* When the WL sub-system returns a physical eraseblock, the physical
* eraseblock is protected from being moved for some "time". For this reason,
@@ -83,6 +83,8 @@
* used. The former state corresponds to the @wl->free tree. The latter state
* is split up on several sub-states:
* o the WL movement is allowed (@wl->used tree);
+ * o the WL movement is disallowed (@wl->erroneous) becouse the PEB is
+ * erroneous - e.g., there was a read error;
* o the WL movement is temporarily prohibited (@wl->pq queue);
* o scrubbing is needed (@wl->scrub tree).
*
@@ -653,7 +655,7 @@ static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
int cancel)
{
- int err, scrubbing = 0, torture = 0, protect = 0;
+ int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0;
struct ubi_wl_entry *e1, *e2;
struct ubi_vid_hdr *vid_hdr;
@@ -769,13 +771,31 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
goto out_not_moved;
}
- if (err == MOVE_CANCEL_BITFLIPS ||
- err == MOVE_TARGET_WR_ERR) {
+ if (err == MOVE_CANCEL_BITFLIPS || err == MOVE_TARGET_WR_ERR ||
+ err == MOVE_TARGET_RD_ERR) {
/* Target PEB bit-flips or write error, torture it */
torture = 1;
goto out_not_moved;
}
+ if (err == MOVE_SOURCE_RD_ERR) {
+ /*
+ * An error happened while reading the source PEB. Do
+ * not switch to R/O mode in this case, and give the
+ * upper layers a possibility to recover from this,
+ * e.g. by unmapping corresponding LEB. Instead, just
+ * put thie PEB to the @ubi->erroneus list to prevent
+ * UBI from trying to move the over and over again.
+ */
+ if (ubi->erroneous_peb_count > ubi->max_erroneous) {
+ ubi_err("too many erroneous eraseblocks (%d)",
+ ubi->erroneous_peb_count);
+ goto out_error;
+ }
+ erroneous = 1;
+ goto out_not_moved;
+ }
+
if (err < 0)
goto out_error;
@@ -832,7 +852,10 @@ out_not_moved:
spin_lock(&ubi->wl_lock);
if (protect)
prot_queue_add(ubi, e1);
- else if (scrubbing)
+ else if (erroneous) {
+ wl_tree_add(e1, &ubi->erroneous);
+ ubi->erroneous_peb_count += 1;
+ } else if (scrubbing)
wl_tree_add(e1, &ubi->scrub);
else
wl_tree_add(e1, &ubi->used);
@@ -1116,6 +1139,13 @@ retry:
} else if (in_wl_tree(e, &ubi->scrub)) {
paranoid_check_in_wl_tree(e, &ubi->scrub);
rb_erase(&e->u.rb, &ubi->scrub);
+ } else if (in_wl_tree(e, &ubi->erroneous)) {
+ paranoid_check_in_wl_tree(e, &ubi->erroneous);
+ rb_erase(&e->u.rb, &ubi->erroneous);
+ ubi->erroneous_peb_count -= 1;
+ ubi_assert(ubi->erroneous_peb_count >= 0);
+ /* Erronious PEBs should be tortured */
+ torture = 1;
} else {
err = prot_queue_del(ubi, e->pnum);
if (err) {
@@ -1364,7 +1394,7 @@ int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
struct ubi_scan_leb *seb, *tmp;
struct ubi_wl_entry *e;
- ubi->used = ubi->free = ubi->scrub = RB_ROOT;
+ ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT;
spin_lock_init(&ubi->wl_lock);
mutex_init(&ubi->move_mutex);
init_rwsem(&ubi->work_sem);
@@ -1502,6 +1532,7 @@ void ubi_wl_close(struct ubi_device *ubi)
cancel_pending(ubi);
protection_queue_destroy(ubi);
tree_destroy(&ubi->used);
+ tree_destroy(&ubi->erroneous);
tree_destroy(&ubi->free);
tree_destroy(&ubi->scrub);
kfree(ubi->lookuptbl);