Skip to contentSkip to conversation

mmc: block: Add single read for 4k sector cards

PATCH Christian Löhle 3 participants 3 messages Jun 28, 2022 – Jun 29, 2022 MMC, SD and SDIO

lore
Share & tools
lore
#0 lore raw
Quoted 6 lines

Cards with 4k native sector size may only be read 4k-aligned,

Quoted 1 line

accommodate for this in the single read recovery and use it.

Thanks for the patch.

Quoted 1 line
Quoted 6 lines
Quoted 2 lines
Fixes 81196976ed946 (mmc: block: Add blk-mq support)

FYI checkpatch says:

WARNING: From:/Signed-off-by: email name mismatch: 'From: "Christian Löhle" <[email protected]>' != 'Signed-off-by: Christian Loehle <[email protected]>'

Will be fixed in my future patches, thanks for the hint.

Quoted 21 lines
Quoted 21 lines
---
 drivers/mmc/core/block.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
drivers/mmc/core/block.c +1 -1
index f4a1281658db..a75a208ce203 100644
@@ -176,7 +176,7 @@ static inline int mmc_blk_part_switch(struct mmc_card *card,
 				      unsigned int part_type);
 static void mmc_blk_rw_rq_prep(struct mmc_queue_req *mqrq,
 			       struct mmc_card *card,
-			       int disable_multi,
+			       int recovery_mode,
 			       struct mmc_queue *mq);
 static void mmc_blk_hsq_req_done(struct mmc_request *mrq);
 
@@ -1302,7 +1302,7 @@ static void mmc_blk_eval_resp_error(struct mmc_blk_request *brq)
 }
 
Quoted 16 lines
Quoted 7 lines
drivers/mmc/core/block.c +1 -1
 static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq,
-			      int disable_multi, bool *do_rel_wr_p,
+			      int recovery_mode, bool *do_rel_wr_p,
 			      bool *do_data_tag_p)
 {
 	struct mmc_blk_data *md = mq->blkdata;

drivers/mmc/core/block.c static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq, · line 1376

Quoted 7 lines
drivers/mmc/core/block.c +2 -2
@@ -1372,8 +1372,8 @@ static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq,
 		 * at a time in order to accurately determine which
 		 * sectors can be read successfully.
 		 */
-		if (disable_multi)
-			brq->data.blocks = 1;
+		if (recovery_mode)
+			brq->data.blocks = mmc_large_sector(card) ? 8 : 1;

I suggest changing to use queue_physical_block_size() here and further below

This part I'm impartial about, not sure if it makes it more readable, hopefully we never have to support another "native sector size" apart from the two. Anyway I will send the next patch with queue_physical_block_size()

Quoted 2 lines
			brq->data.blocks = queue_physical_block_size(req->q) >> SECTOR_SHIFT;

Do we want to switch to SECTOR_SHIFT instead of 9? So far SECTOR_SHIFT is not used at all in mmc core. If so I would go ahead and change all the others in another patch:
queue.c:187: q->limits.discard_granularity = card->pref_erase << 9;
core.c:103: data->bytes_xfered = (prandom_u32() % (data->bytes_xfered >> 9)) << 9;
mmc.c:792:MMC_DEV_ATTR(erase_size, "%u\n", card->erase_size << 9); mmc.c:793:MMC_DEV_ATTR(preferred_erase_size, "%u\n", card->pref_erase << 9);
mmc_test.c:1557: sz = (unsigned long)test->card->pref_erase << 9;
mmc_test.c:1570: t->max_tfr = test->card->host->max_blk_count << 9;
mmc_test.c:2461: if (repeat_cmd && (t->blocks + 1) << 9 > t->max_tfr)
sd.c:707:MMC_DEV_ATTR(erase_size, "%u\n", card->erase_size << 9); sd.c:708:MMC_DEV_ATTR(preferred_erase_size, "%u\n", card->pref_erase << 9);
block.c:1417: int i, data_size = brq->data.blocks << 9;
block.c:1851: brq->data.bytes_xfered = blocks << 9;

Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz Managing Director: Dr. Jan Peter Berns. Commercial register of local courts: Freiburg HRB381782

#1 lore raw
Quoted 53 lines · Christian Löhle
Quoted 6 lines

Cards with 4k native sector size may only be read 4k-aligned,

Quoted 1 line

accommodate for this in the single read recovery and use it.

Thanks for the patch.

Quoted 1 line
Quoted 6 lines
Quoted 2 lines
Fixes 81196976ed946 (mmc: block: Add blk-mq support)

FYI checkpatch says:

WARNING: From:/Signed-off-by: email name mismatch: 'From: "Christian Löhle" <[email protected]>' != 'Signed-off-by: Christian Loehle <[email protected]>'

Will be fixed in my future patches, thanks for the hint.

Quoted 21 lines
Quoted 21 lines
---
 drivers/mmc/core/block.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
drivers/mmc/core/block.c +1 -1
index f4a1281658db..a75a208ce203 100644
@@ -176,7 +176,7 @@ static inline int mmc_blk_part_switch(struct mmc_card *card,
 				      unsigned int part_type);
 static void mmc_blk_rw_rq_prep(struct mmc_queue_req *mqrq,
 			       struct mmc_card *card,
-			       int disable_multi,
+			       int recovery_mode,
 			       struct mmc_queue *mq);
 static void mmc_blk_hsq_req_done(struct mmc_request *mrq);
 
@@ -1302,7 +1302,7 @@ static void mmc_blk_eval_resp_error(struct mmc_blk_request *brq)
 }
 
Quoted 16 lines
Quoted 7 lines
drivers/mmc/core/block.c +1 -1
 static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq,
-			      int disable_multi, bool *do_rel_wr_p,
+			      int recovery_mode, bool *do_rel_wr_p,
 			      bool *do_data_tag_p)
 {
 	struct mmc_blk_data *md = mq->blkdata;

drivers/mmc/core/block.c static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq, · line 1376

Quoted 7 lines
drivers/mmc/core/block.c +2 -2
@@ -1372,8 +1372,8 @@ static void mmc_blk_data_prep(struct mmc_queue *mq, struct mmc_queue_req *mqrq,
 		 * at a time in order to accurately determine which
 		 * sectors can be read successfully.
 		 */
-		if (disable_multi)
-			brq->data.blocks = 1;
+		if (recovery_mode)
+			brq->data.blocks = mmc_large_sector(card) ? 8 : 1;

I suggest changing to use queue_physical_block_size() here and further below

Quoted 7 lines

This part I'm impartial about, not sure if it makes it more readable, hopefully we never have to support another "native sector size" apart from the two. Anyway I will send the next patch with queue_physical_block_size()

Quoted 2 lines
			brq->data.blocks = queue_physical_block_size(req->q) >> SECTOR_SHIFT;

Do we want to switch to SECTOR_SHIFT instead of 9? So far SECTOR_SHIFT is not used at all in mmc core.

I guess '9' is more consistent

Quoted 19 lines

If so I would go ahead and change all the others in another patch:
queue.c:187: q->limits.discard_granularity = card->pref_erase << 9;
core.c:103: data->bytes_xfered = (prandom_u32() % (data->bytes_xfered >> 9)) << 9;
mmc.c:792:MMC_DEV_ATTR(erase_size, "%u\n", card->erase_size << 9); mmc.c:793:MMC_DEV_ATTR(preferred_erase_size, "%u\n", card->pref_erase << 9);
mmc_test.c:1557: sz = (unsigned long)test->card->pref_erase << 9;
mmc_test.c:1570: t->max_tfr = test->card->host->max_blk_count << 9;
mmc_test.c:2461: if (repeat_cmd && (t->blocks + 1) << 9 > t->max_tfr)
sd.c:707:MMC_DEV_ATTR(erase_size, "%u\n", card->erase_size << 9); sd.c:708:MMC_DEV_ATTR(preferred_erase_size, "%u\n", card->pref_erase << 9);
block.c:1417: int i, data_size = brq->data.blocks << 9;
block.c:1851: brq->data.bytes_xfered = blocks << 9;

Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz Managing Director: Dr. Jan Peter Berns. Commercial register of local courts: Freiburg HRB381782

#2 lore raw

...

Quoted 6 lines
Quoted 3 lines
Quoted 1 line
			brq->data.blocks = queue_physical_block_size(req->q) >> SECTOR_SHIFT;

Do we want to switch to SECTOR_SHIFT instead of 9? So far SECTOR_SHIFT is not used at all in mmc

core.

I guess '9' is more consistent

You can just multiply/divide by 512u (probably SECTOR_SIZE). (In some senses 512u is actually better!)
More obvious still and the compiler will generate a shift anyway.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)