Skip to contentSkip to conversation

mtd: fix hang-up in cfi erase and read contention

PATCH Tadashi Abe 4 participants 5 messages May 19, 2011 – May 25, 2011 Memory technology devices (MTD)

lore
Share & tools
lore
#0 lore raw

cfi erase command hangs up when erase and read contention occurs. If read runs at the same address as erase operation, read issues Erase-Suspend via get_chip() and the erase goes into sleep in wait queue. But in this case, read operation exits by time-out without waking it up.

I think the other variants (0001, 0020 and lpddr) have the same problem too. Tested and verified the patch only on CFI-0002 flash, though.

---
 drivers/mtd/chips/cfi_cmdset_0001.c |    9 +++------
 drivers/mtd/chips/cfi_cmdset_0002.c |    4 +---
 drivers/mtd/chips/cfi_cmdset_0020.c |    1 +
 drivers/mtd/lpddr/lpddr_cmds.c      |    7 +------
 4 files changed, 6 insertions(+), 15 deletions(-)
drivers/mtd/chips/cfi_cmdset_0001.c +3 -6
index 09cb7c8..121be02 100644
@@ -812,12 +812,9 @@ static int chip_ready (struct map_info *map, struct flchip *chip, unsigned long
 			        break;
 
 			if (time_after(jiffies, timeo)) {
-				/* Urgh. Resume and pretend we weren't here.  */
-				map_write(map, CMD(0xd0), adr);
-				/* Make sure we're in 'read status' mode if it had finished */
-				map_write(map, CMD(0x70), adr);
-				chip->state = FL_ERASING;
-				chip->oldstate = FL_READY;
+				/* Urgh. Resume and pretend we weren't here.
+				 * Make sure we're in 'read status' mode if it had finished */
+				put_chip(map, chip, adr);
 				printk(KERN_ERR "%s: Chip not ready after erase "
 				       "suspended: status = 0x%lx\n", map->name, status.x[0]);
 				return -EIO;
drivers/mtd/chips/cfi_cmdset_0002.c +1 -3
index 0b49266..e97e697 100644
@@ -710,9 +710,7 @@ static int get_chip(struct map_info *map, struct flchip *chip, unsigned long adr
 				 * there was an error (so leave the erase
 				 * routine to recover from it) or we trying to
 				 * use the erase-in-progress sector. */
-				map_write(map, cfi->sector_erase_cmd, chip->in_progress_block_addr);
-				chip->state = FL_ERASING;
-				chip->oldstate = FL_READY;
+				put_chip(map, chip, adr);
 				printk(KERN_ERR "MTD %s(): chip not ready after erase suspend\n", __func__);
 				return -EIO;
 			}
drivers/mtd/chips/cfi_cmdset_0020.c +1 -0
index ed56ad3..179814a 100644
@@ -296,6 +296,7 @@ static inline int do_read_onechip(struct map_info *map, struct flchip *chip, lof
 				/* make sure we're in 'read status' mode */
 				map_write(map, CMD(0x70), cmd_addr);
 				chip->state = FL_ERASING;
+				wake_up(&chip->wq);
 				mutex_unlock(&chip->mutex);
 				printk(KERN_ERR "Chip not ready after erase "
 				       "suspended: status = 0x%lx\n", status.x[0]);
drivers/mtd/lpddr/lpddr_cmds.c +1 -6
index 1267992..16dcd1c 100644
@@ -313,12 +313,7 @@ static int chip_ready(struct map_info *map, struct flchip *chip, int mode)
 		if (ret) {
 			/* Oops. something got wrong. */
 			/* Resume and pretend we weren't here.  */
-			map_write(map, CMD(LPDDR_RESUME),
-				map->pfow_base + PFOW_COMMAND_CODE);
-			map_write(map, CMD(LPDDR_START_EXECUTION),
-				map->pfow_base + PFOW_COMMAND_EXECUTE);
-			chip->state = FL_ERASING;
-			chip->oldstate = FL_READY;
+			put_chip(map, chip);
 			printk(KERN_ERR "%s: suspend operation failed."
 					"State may be wrong \n", map->name);
 			return -EIO;
Signature · 1 line
1.7.4.4
#1 lore raw
Quoted 11 lines

From: Tadashi Abe <[email protected]>

cfi erase command hangs up when erase and read contention occurs. If read runs at the same address as erase operation, read issues Erase-Suspend via get_chip() and the erase goes into sleep in wait queue. But in this case, read operation exits by time-out without waking it up.

I think the other variants (0001, 0020 and lpddr) have the same problem too. Tested and verified the patch only on CFI-0002 flash, though.

This looks good for cfi_cmdset_0001.c so:

#2 lore raw
In reply to Tadashi Abe
Quoted 9 lines · Tadashi Abe

cfi erase command hangs up when erase and read contention occurs. If read runs at the same address as erase operation, read issues Erase-Suspend via get_chip() and the erase goes into sleep in wait queue. But in this case, read operation exits by time-out without waking it up.

I think the other variants (0001, 0020 and lpddr) have the same problem too. Tested and verified the patch only on CFI-0002 flash, though.

Pushed both cfi patches to l2-mtd-2.6.git, thanks.

Signature · 2 lines
Best Regards,
Artem Bityutskiy (Артём Битюцкий)
#3 lore raw
In reply to Tadashi Abe

drivers/mtd/chips/cfi_cmdset_0001.c static int chip_ready (struct map_info *map, struct flchip *chip, unsigned long · line 817

Quoted 13 lines · Tadashi Abe
drivers/mtd/chips/cfi_cmdset_0001.c +3 -6
@@ -812,12 +812,9 @@ static int chip_ready (struct map_info *map, struct flchip *chip, unsigned long
 			        break;
 
 			if (time_after(jiffies, timeo)) {
-				/* Urgh. Resume and pretend we weren't here.  */
-				map_write(map, CMD(0xd0), adr);
-				/* Make sure we're in 'read status' mode if it had finished */
-				map_write(map, CMD(0x70), adr);
-				chip->state = FL_ERASING;
-				chip->oldstate = FL_READY;
+				/* Urgh. Resume and pretend we weren't here.
+				 * Make sure we're in 'read status' mode if it had finished */
+				put_chip(map, chip, adr);

I'm confused. You've removed the code which ensures that the chip is in a known state.... why?

Signature · 1 line
dwmw2
#4 lore raw
Quoted 2 lines · David Woodhouse

I'm confused. You've removed the code which ensures that the chip is in a known state.... why?

Oh, because put_chip() does it :)

Signature · 1 line
dwmw2