Skip to contentSkip to conversation

mmc: limit block count for multi-block write commands

PATCH New Lei Wen 4 participants 16 messages Sep 7, 2010 – Sep 13, 2010 MMC, SD and SDIO

lore
Share & tools
lore

Lei Wen

author
#0 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

---
 drivers/mmc/mmc.c |   59 ++++++++++++++++++++++++++++++++++------------------
 1 files changed, 38 insertions(+), 21 deletions(-)
drivers/mmc/mmc.c +38 -21
index cf4ea16..5cc1904 100644
@@ -77,29 +77,14 @@ struct mmc *find_mmc_device(int dev_num)
 	return NULL;
 }
 
-static ulong
-mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
+int mmc_write_block(struct mmc *mmc, void *src, ulong start, uint blocknum)
 {
-	struct mmc_cmd cmd;
-	struct mmc_data data;
-	int err;
-	int stoperr = 0;
-	struct mmc *mmc = find_mmc_device(dev_num);
 	int blklen;
-
-	if (!mmc)
-		return -1;
-
 	blklen = mmc->write_bl_len;
 
-	err = mmc_set_blocklen(mmc, mmc->write_bl_len);
-
-	if (err) {
-		printf("set write bl len failed\n\r");
-		return err;
-	}
-
-	if (blkcnt > 1)
+	BUG_ON(blklen * blocknum > 524288);
+	BUG_ON(blocknum > 65535);
+	if (blocknum > 1)
 		cmd.cmdidx = MMC_CMD_WRITE_MULTIPLE_BLOCK;
 	else
 		cmd.cmdidx = MMC_CMD_WRITE_SINGLE_BLOCK;
@@ -113,7 +98,7 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 	cmd.flags = 0;
 
 	data.src = src;
-	data.blocks = blkcnt;
+	data.blocks = blocknum;
 	data.blocksize = blklen;
 	data.flags = MMC_DATA_WRITE;
 
@@ -124,7 +109,7 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 		return err;
 	}
 
-	if (blkcnt > 1) {
+	if (blocknum > 1) {
 		cmd.cmdidx = MMC_CMD_STOP_TRANSMISSION;
 		cmd.cmdarg = 0;
 		cmd.resp_type = MMC_RSP_R1b;
@@ -132,6 +117,38 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 		stoperr = mmc_send_cmd(mmc, &cmd, NULL);
 	}
 
+	return blocknum;
+}
+
+static ulong
+mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
+{
+	struct mmc_cmd cmd;
+	struct mmc_data data;
+	int err;
+	int stoperr = 0;
+	struct mmc *mmc = find_mmc_device(dev_num);
+	unsigned int max;
+	lbaint_t tmp = blkcnt, cur;
+
+	if (!mmc)
+		return -1;
+
+	err = mmc_set_blocklen(mmc, mmc->write_bl_len);
+	if (err) {
+		printf("set write bl len failed\n\r");
+		return err;
+	}
+
+	max = 524288 / mmc->write_bl_len;
+	do {
+		cur = (tmp > max) ? max : tmp;
+		if(mmc_write_block(mmc, src, start, cur) != cur)
+			return -1;
+		tmp -= cur;
+		start += cur;
+		src += cur * mmc->write_bl_len;
+	} while (tmp > 0);
 	return blkcnt;
 }
 
Signature · 1 line
1.7.0.4
author
#1 lore raw

PATCH 2/3 mmc: add boundary check for mmc operation

---
 drivers/mmc/mmc.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
drivers/mmc/mmc.c +8 -0
index 5cc1904..9a50b2f 100644
@@ -134,6 +134,10 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 	if (!mmc)
 		return -1;
 
+	if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+		printf("\noperation excceed mmc boudary..\n");
+		return 0;
+	}
 	err = mmc_set_blocklen(mmc, mmc->write_bl_len);
 	if (err) {
 		printf("set write bl len failed\n\r");
@@ -236,6 +240,10 @@ static ulong mmc_bread(int dev_num, ulong start, lbaint_t blkcnt, void *dst)
 	if (!mmc)
 		return 0;
 
+	if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+		printf("\noperation excceed mmc boudary..\n");
+		return 0;
+	}
 	/* We always do full block reads from the card */
 	err = mmc_set_blocklen(mmc, mmc->read_bl_len);
 
Signature · 1 line
1.7.0.4
#2 lore raw

PATCH 2/3 mmc: add boundary check for mmc operation

Dear Lei Wen,

Quoted 10 lines · In message <[email protected]> you wrote:
---
 drivers/mmc/mmc.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
drivers/mmc/mmc.c +0 -0
index 5cc1904..9a50b2f 100644

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 138

Quoted 5 lines
drivers/mmc/mmc.c +2 -0
@@ -134,6 +134,10 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 	if (!mmc)
 		return -1;
 
+	if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+		printf("\noperation excceed mmc boudary..\n");

No initial newline, please.

Please fix typos: exceed, boundary.

Use puts() instead of printf() for messages that do not need formatting.

Hm... Maybe you could change the message so it is more informatiove, i. e. tell the user which boundary was exceeded, and what caused this condition - is this a user error or a problem in the code?

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 143

Quoted 3 lines
drivers/mmc/mmc.c +0 -0
 	err = mmc_set_blocklen(mmc, mmc->write_bl_len);
 	if (err) {
 		printf("set write bl len failed\n\r");

While you are at it: please fix this message, too. Remove the '\r', and replace the "bl len" with something the user can understand.

drivers/mmc/mmc.c static ulong mmc_bread(int dev_num, ulong start, lbaint_t blkcnt, void *dst) · line 244

Quoted 6 lines
drivers/mmc/mmc.c +2 -0
@@ -236,6 +240,10 @@ static ulong mmc_bread(int dev_num, ulong start, lbaint_t blkcnt, void *dst)
 	if (!mmc)
 		return 0;
 
+	if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+		printf("\noperation excceed mmc boudary..\n");

See above.

Hm... This is repeated code. Factor it out?

Best regards,

Wolfgang Denk

Signature · 4 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
There's an old proverb that says just about whatever you want it to.
#3 lore raw

PATCH 2/3 mmc: add boundary check for mmc operation

Quoted 16 lines · Wolfgang Denk

Dear Lei Wen,

Quoted 13 lines · In message <[email protected]> you wrote:

---
?drivers/mmc/mmc.c | ? ?8 ++++++++
?1 files changed, 8 insertions(+), 0 deletions(-)

drivers/mmc/mmc.c +0 -0
index 5cc1904..9a50b2f 100644
@@ -134,6 +134,10 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 ? ? ? if (!mmc)
 ? ? ? ? ? ? ? return -1;
 
Quoted 4 lines

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 138

Quoted 2 lines
drivers/mmc/mmc.c +2 -0
+ ? ? if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+ ? ? ? ? ? ? printf("\noperation excceed mmc boudary..\n");

No initial newline, please.

You mean add additional line or just reduce a line here?...

Quoted 2 lines

Please fix typos: exceed, boundary.

Ok... Sorry for that...

Quoted 3 lines

Use puts() instead of printf() for messages that do not need formatting.

Em, ok.

Quoted 4 lines

Hm... Maybe you could change the message so it is more informatiove, i. e. tell the user which boundary was exceeded, and what caused this condition - is this a user error or a problem in the code?

Give error message for each judgment of the two?

Quoted 12 lines

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 141

Quoted 4 lines
drivers/mmc/mmc.c +0 -0
 ? ? ? err = mmc_set_blocklen(mmc, mmc->write_bl_len);
 ? ? ? if (err) {
 ? ? ? ? ? ? ? printf("set write bl len failed\n\r");

While you are at it: please fix this message, too. Remove the '\r', and replace the "bl len" with something the user can understand.

Quoted 4 lines
drivers/mmc/mmc.c +0 -0
@@ -236,6 +240,10 @@ static ulong mmc_bread(int dev_num, ulong start, lbaint_t blkcnt, void *dst)
 ? ? ? if (!mmc)
 ? ? ? ? ? ? ? return 0;
 
Quoted 6 lines

drivers/mmc/mmc.c static ulong mmc_bread(int dev_num, ulong start, lbaint_t blkcnt, void *dst) · line 244

Quoted 2 lines
drivers/mmc/mmc.c +2 -0
+ ? ? if (start > mmc->block_dev.lba || (start + blkcnt) > mmc->block_dev.lba) {
+ ? ? ? ? ? ? printf("\noperation excceed mmc boudary..\n");

See above.

Hm... This is repeated code. Factor it out?

It is so hard to factor... There is no common up layer calling, or down layer calling to add this check...

Quoted 12 lines

Best regards,

Wolfgang Denk

--
DENX Software Engineering GmbH, ? ? MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de There's an old proverb that says just about whatever you want it to.

#4 lore raw

PATCH 2/3 mmc: add boundary check for mmc operation

Dear Lei Wen,

Quoted 7 lines · In message <[email protected]> you…
Quoted 5 lines

Review

Quoted 3 lines
(quoted diff) +2 -0
+       if (start > mmc->block_dev.lba || (start + blkcnt) mmc->block_dev.lba) {
+                   printf("\noperation excceed mmc boudary..\n");

No initial newline, please.

You mean add additional line or just reduce a line here?...

I mean change

	printf("\noperation ...");
into
	printf("operation ...");
Quoted 5 lines
Quoted 3 lines

Hm... Maybe you could change the message so it is more informatiove, i. e. tell the user which boundary was exceeded, and what caused this condition - is this a user error or a problem in the code?

Give error message for each judgment of the two?

Not really needed, but you might print the legal range and the current value in the error message.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Time is a drug. Too much of it kills you.
                                      - Terry Pratchett, _Small Gods_
#5 lore raw

PATCH 2/3 mmc: add boundary check for mmc operation

Fix it in new version of patch set.

Thanks,
Lei

Quoted 37 lines · Wolfgang Denk

Dear Lei Wen,

Quoted 7 lines · In message <[email protected]> you…
Quoted 5 lines
Quoted 3 lines

+ ? ? ? if (start > mmc->block_dev.lba || (start + blkcnt) mmc->block_dev.lba) {
+ ? ? ? ? ? ? ? ? ? printf("\noperation excceed mmc boudary..\n");

No initial newline, please.

You mean add additional line or just reduce a line here?...

I mean change

? ? ? ?printf("\noperation ...");
into
? ? ? ?printf("operation ...");

Quoted 5 lines
Quoted 3 lines

Hm... Maybe you could change the message so it is more informatiove, i. e. tell the user which boundary was exceeded, and what caused this condition - is this a user error or a problem in the code?

Give error message for each judgment of the two?

Not really needed, but you might print the legal range and the current value in the error message.

Best regards,

Wolfgang Denk

--
DENX Software Engineering GmbH, ? ? MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de Time is a drug. Too much of it kills you.
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?- Terry Pratchett, _Small Gods_

author
#6 lore raw

PATCH 3/3 MMC: print out avaible partition table

In reply to Lei Wen
---
 common/cmd_mmc.c |   20 ++++++++++++++++++++
 disk/part.c      |    3 +++
 2 files changed, 23 insertions(+), 0 deletions(-)
common/cmd_mmc.c +20 -0
index c0b30d8..af82984 100644
@@ -154,6 +154,25 @@ int do_mmcops(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 			mmc_init(mmc);
 
 			return 0;
+		} else if (strncmp(argv[1], "part", 4) == 0) {
+			int dev = simple_strtoul(argv[2], NULL, 10);
+			block_dev_desc_t *mmc_dev;
+			struct mmc *mmc = find_mmc_device(dev);
+
+			if (!mmc)
+				return 1;
+			mmc_init(mmc);
+			mmc_dev = mmc_get_dev(dev);
+			if (mmc_dev != NULL &&
+					mmc_dev->type != DEV_TYPE_UNKNOWN) {
+				rc = 1;
+				print_part(mmc_dev);
+			}
+			if (!rc) {
+				printf("\nno mmc devices available\n");
+				return 1;
+			}
+			return 0;
 		}
 
 	case 0:
@@ -230,5 +249,6 @@ U_BOOT_CMD(
 	"read <device num> addr blk# cnt\n"
 	"mmc write <device num> addr blk# cnt\n"
 	"mmc rescan <device num>\n"
+	"mmc part <device num>- lists avaiable partition on mmc\n"
 	"mmc list - lists available devices");
 #endif
disk/part.c +3 -0
index 3ba88c7..1806fe6 100644
@@ -364,6 +364,9 @@ static void print_part_header (const char *type, block_dev_desc_t * dev_desc)
 	case IF_TYPE_DOC:
 		puts ("DOC");
 		break;
+	case IF_TYPE_MMC:
+		puts ("MMC");
+		break;
 	default:
 		puts ("UNKNOWN");
 		break;
Signature · 1 line
1.7.0.4
#7 lore raw

PATCH 3/3 MMC: print out avaible partition table

Dear Lei Wen,

Quoted 11 lines · In message <[email protected]> you wrote:
---
 common/cmd_mmc.c |   20 ++++++++++++++++++++
 disk/part.c      |    3 +++
 2 files changed, 23 insertions(+), 0 deletions(-)
common/cmd_mmc.c +0 -0
index c0b30d8..af82984 100644

common/cmd_mmc.c int do_mmcops(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[]) · line 163

Quoted 10 lines
common/cmd_mmc.c +7 -0
@@ -154,6 +154,25 @@ int do_mmcops(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 			mmc_init(mmc);
 
 			return 0;
+		} else if (strncmp(argv[1], "part", 4) == 0) {
+			int dev = simple_strtoul(argv[2], NULL, 10);
+			block_dev_desc_t *mmc_dev;
+			struct mmc *mmc = find_mmc_device(dev);
+
+			if (!mmc)
+				return 1;

Please print error message in case of errors.

common/cmd_mmc.c int do_mmcops(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[]) · line 169

Quoted 6 lines
common/cmd_mmc.c +6 -0
+			mmc_init(mmc);
+			mmc_dev = mmc_get_dev(dev);
+			if (mmc_dev != NULL &&
+					mmc_dev->type != DEV_TYPE_UNKNOWN) {
+				rc = 1;
+				print_part(mmc_dev);

Do we need to set rc at all? Just do a "return 0" here.

common/cmd_mmc.c int do_mmcops(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[]) · line 173

Quoted 4 lines
common/cmd_mmc.c +4 -0
+			}
+			if (!rc) {
+				printf("\nno mmc devices available\n");
+				return 1;

No leading newline, please. And use puts() for strings that do not need formatting.

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Youth doesn't excuse everything.
	-- Dr. Janice Lester (in Kirk's body), "Turnabout Intruder",
	   stardate 5928.5.
#8 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

In reply to Lei Wen

Dear Lei Wen,

drivers/mmc/mmc.c struct mmc *find_mmc_device(int dev_num) · line 86

Quoted 7 lines · In message <[email protected]> you wrote:
---
 drivers/mmc/mmc.c |   59 ++++++++++++++++++++++++++++++++++------------------
 1 files changed, 38 insertions(+), 21 deletions(-)
drivers/mmc/mmc.c +2 -0
+	BUG_ON(blklen * blocknum > 524288);
+	BUG_ON(blocknum > 65535);

Please add a comment that explains where these magic numbers are coming from.

And are you sure that these are conditions where the system should hang? I think this is not an approrpiate way to handle such errors. Just failing the command should be enough.

drivers/mmc/mmc.c struct mmc *find_mmc_device(int dev_num) · line 87

Quoted 1 line
drivers/mmc/mmc.c +1 -0
+	if (blocknum > 1)

I already complained that I consider "blocknum" to be inappropriately named. It is not a block number (= number of a block), but a block count.

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 101

Quoted 2 lines
drivers/mmc/mmc.c +1 -1
-	data.blocks = blkcnt;
+	data.blocks = blocknum;

The previously used name was way better!

Quoted 13 lines
drivers/mmc/mmc.c +1 -1
 	data.blocksize = blklen;
 	data.flags = MMC_DATA_WRITE;
@@ -124,7 +109,7 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 		return err;
 	}
 
-	if (blkcnt > 1) {
+	if (blocknum > 1) {
 		cmd.cmdidx = MMC_CMD_STOP_TRANSMISSION;
 		cmd.cmdarg = 0;
 		cmd.resp_type = MMC_RSP_R1b;

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 119

Quoted 3 lines
drivers/mmc/mmc.c +1 -0
@@ -132,6 +117,38 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
 		stoperr = mmc_send_cmd(mmc, &cmd, NULL);
 	}
+	max = 524288 / mmc->write_bl_len;

Please explain where these magic number is coming from. Eventually use #defines.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Commitment, n.:      Commitment can be illustrated by a breakfast
of ham and eggs. The chicken was involved, the pig was committed.
#9 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

Quoted 36 lines · Wolfgang Denk

Dear Lei Wen,

Review

Quoted 7 lines · In message<[email protected]> you wrote:
---
  drivers/mmc/mmc.c |   59 ++++++++++++++++++++++++++++++++++------------------
  1 files changed, 38 insertions(+), 21 deletions(-)
(quoted diff) +2 -0
+	BUG_ON(blklen * blocknum>  524288);
+	BUG_ON(blocknum>  65535);

Please add a comment that explains where these magic numbers are coming from.

And are you sure that these are conditions where the system should hang? I think this is not an approrpiate way to handle such errors. Just failing the command should be enough.

Review

Quoted 1 line
(quoted diff) +1 -0
+	if (blocknum>  1)

I already complained that I consider "blocknum" to be inappropriately named. It is not a block number (= number of a block), but a block count.

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 101

Quoted 2 lines
drivers/mmc/mmc.c +1 -1
-	data.blocks = blkcnt;
+	data.blocks = blocknum;

The previously used name was way better!

Quoted 7 lines
  	data.blocksize = blklen;
  	data.flags = MMC_DATA_WRITE;
drivers/mmc/mmc.c +0 -0
@@ -124,7 +109,7 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
  		return err;
  	}
 
Quoted 12 lines

drivers/mmc/mmc.c mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src) · line 119

Quoted 9 lines
drivers/mmc/mmc.c +2 -1
-	if (blkcnt>  1) {
+	if (blocknum>  1) {
  		cmd.cmdidx = MMC_CMD_STOP_TRANSMISSION;
  		cmd.cmdarg = 0;
  		cmd.resp_type = MMC_RSP_R1b;
@@ -132,6 +117,38 @@ mmc_bwrite(int dev_num, ulong start, lbaint_t blkcnt, const void*src)
  		stoperr = mmc_send_cmd(mmc,&cmd, NULL);
  	}
+	max = 524288 / mmc->write_bl_len;

Please explain where these magic number is coming from. Eventually use #defines.

Besides that my remarks to yesterdays patch of yours are still valid:
Those "magic numbers" are due to a specific hardware controller/limitation and not any SD/MMC card limitation.

And which hardware driver are you using this with?
Maybe some context would help.
Maybe the splitting could be done in the hardware driver and not in this generic part?

Reinhard

#10 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

Dear Reinhard Meyer,

Quoted 4 lines · In message <[email protected]> you wrote:

Besides that my remarks to yesterdays patch of yours are still valid:
Those "magic numbers" are due to a specific hardware controller/limitation and not any SD/MMC card limitation.

Correct.

Quoted 4 lines

And which hardware driver are you using this with?
Maybe some context would help.
Maybe the splitting could be done in the hardware driver and not in this generic part?

That would indeed be a better choice.

Best regards,

Wolfgang Denk

Signature · 7 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"Evil does seek to maintain power by suppressing the truth."
"Or by misleading the innocent."
	-- Spock and McCoy, "And The Children Shall Lead",
	   stardate 5029.5.
#11 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

In reply to Reinhard Meyer

Hi Reinhard,

Quoted 9 lines

Besides that my remarks to yesterdays patch of yours are still valid:
Those "magic numbers" are due to a specific hardware controller/limitation and not any SD/MMC card limitation.

And which hardware driver are you using this with?
Maybe some context would help.
Maybe the splitting could be done in the hardware driver and not in this generic part?

Certainly those limitation could not be put at the generic part. But for the common limitation from the spec,
I believe put there is a better choice. I mention for the sd host controller standard spec has such limitation for a maximum blocks at one go.
You could refer to the pdf doc at:
http://www.sdcard.org/developers/tech/host_controller/simple_spec/Simplified_SD_Host_Controller_Spec.pdf
The register table at page 20 shows the block count register is only 16 bit width.

For the 524288 magic number, 512k, I just follow the Linux way, that is don't let the driver to bother to handle the dma boundary if it use sdma instead PIO.

So it is not which hardware's question.

Best regards,
Lei

#12 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

Dear Lei Wen,

Quoted 12 lines

Hi Reinhard,

Quoted 9 lines

Besides that my remarks to yesterdays patch of yours are still valid:
Those "magic numbers" are due to a specific hardware controller/limitation and not any SD/MMC card limitation.

And which hardware driver are you using this with?
Maybe some context would help.
Maybe the splitting could be done in the hardware driver and not in this generic part?

Quoted 9 lines

Certainly those limitation could not be put at the generic part. But for the common limitation from the spec,
I believe put there is a better choice. I mention for the sd host controller standard spec has such limitation for a maximum blocks at one go.
You could refer to the pdf doc at:
http://www.sdcard.org/developers/tech/host_controller/simple_spec/Simplified_SD_Host_Controller_Spec.pdf
The register table at page 20 shows the block count register is only 16 bit width.

This is now going in circles...
You keep repeating the same Tantra over and over. You should by now have realized that this spec gives an EXAMPLE of how a host controller might be realized. For obvious reasons most designers will NOT follow that SD-Card specification and might even call their controller a multi-media-card interface.

So I repeat again: I am sure that many embedded systems do NOT have such a "Simplified_SD_Host_Controller" that has a funny 16 Bit Register in its hardware.

Quoted 3 lines

For the 524288 magic number, 512k, I just follow the Linux way, that is don't let the driver to bother to handle the dma boundary if it use sdma instead PIO.

And again: I am sure most hardware does not have a funny 512k boundary for DMA operations either. And if, would that boundary not be at a divisible by 512k memory address (page size maybe) and not at a length of 512k?

Quoted 2 lines

So it is not which hardware's question.

It IS hardware related.

However if agreement exists that the commom MMC code shall limit its requests to drivers to the most limiting hardware design... be it.

Reinhard

#13 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

Hi Reinhard,

Quoted 30 lines · Reinhard Meyer

Dear Lei Wen,

Quoted 12 lines

Hi Reinhard,

Quoted 9 lines

Besides that my remarks to yesterdays patch of yours are still valid:
Those "magic numbers" are due to a specific hardware controller/limitation and not any SD/MMC card limitation.

And which hardware driver are you using this with?
Maybe some context would help.
Maybe the splitting could be done in the hardware driver and not in this generic part?

Quoted 9 lines

Certainly those limitation could not be put at the generic part. But for the common limitation from the spec,
I believe put there is a better choice. I mention for the sd host controller standard spec has such limitation for a maximum blocks at one go.
You could refer to the pdf doc at:
http://www.sdcard.org/developers/tech/host_controller/simple_spec/Simplified_SD_Host_Controller_Spec.pdf
The register table at page 20 shows the block count register is only 16 bit width.

This is now going in circles...
You keep repeating the same Tantra over and over. You should by now have realized that this spec gives an EXAMPLE of how a host controller might be realized. For obvious reasons most designers will NOT follow that SD-Card specification and might even call their controller a multi-media-card interface.

Quoted 3 lines

So I repeat again: I am sure that many embedded systems do NOT have such a "Simplified_SD_Host_Controller" that has a funny 16 Bit Register in its hardware.

I also agree this discussion should not going into a endless circle... You says that the sd spec is just example, and mention the mmc. But you should not forget mmc&sd card are often designed together, so designer would
choose the sd register layout to get the sd&mmc both works.

If designer don't follow the spec, it is obvious allowable, but this one just should be a special case.
I believe most others would like to follow the spec directly.

Maybe we could let somebody else to show their choice...

Quoted 5 lines
Quoted 4 lines

For the 524288 magic number, 512k, I just follow the Linux way, that is don't let the driver to bother to handle the dma boundary if it use sdma instead PIO.

Quoted 8 lines

And again: I am sure most hardware does not have a funny 512k boundary for DMA operations either. And if, would that boundary not be at a divisible by 512k memory address (page size maybe) and not at a length of 512k?

Quoted 2 lines

So it is not which hardware's question.

It IS hardware related.

Yes, it is hardware related, but I mean not a special hardware...

Quoted 6 lines

However if agreement exists that the commom MMC code shall limit its requests to drivers to the most limiting hardware design... be it.

Reinhard

Best regards,
Lei

Quoted 1 line
#14 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

Dear Lei Wen,

Quoted 9 lines · In message <[email protected]> you…

You says that the sd spec is just example, and mention the mmc. But you should not forget mmc&sd card are often designed together, so designer would
choose the sd register layout to get the sd&mmc both works.

If designer don't follow the spec, it is obvious allowable, but this one just should be a special case.
I believe most others would like to follow the spec directly.

As Reinhard pointed out, the document you quoted is the SD Host Controller Simplified Specification, but we are talking about MMC code here, which is a bit more generic. SD Host Controller limitations should be handled in the SD Host Controller driver code, not in the MMC generic code.

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

For the 524288 magic number, 512k, I just follow the Linux way, that is don't let the driver to bother to handle the dma boundary if it use sdma instead PIO.

And again: I am sure most hardware does not have a funny 512k boundary for DMA operations either. And if, would that boundary not be at a divisible by 512k memory address (page size maybe) and not at a length of 512k?

You did not reply to this argument. Are you really sure the limitation is on the transfer size?

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Any sufficiently advanced  technology  is  indistinguishable  from  a
rigged demo.                              - Andy Finkel, computer guy
#15 lore raw

PATCH 1/3 mmc: limit block count for multi-block write commands

In reply to Reinhard Meyer

Dear Reinhard Meyer,

Quoted 7 lines · In message <[email protected]> you wrote:
Quoted 2 lines

So it is not which hardware's question.

It IS hardware related.

However if agreement exists that the commom MMC code shall limit its requests to drivers to the most limiting hardware design... be it.

No such agreement exists.

Your arguments are well-heard, and from what I've seen so far I am not really happy with adding this patch either.

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
No one wants war.
	-- Kirk, "Errand of Mercy", stardate 3201.7