On eMMC device, the erase_grp_size > 1 so the address and size for the erase block command in env/mmc.c can be unaligned on erase group size and some strange trace occurs and the result is not guarantee by MMC devices.

The SD-Card behavior doesn't change as erase_grp_size = 1 for SD-Card.

For example, on eMMC present on STM32MP15C-EV1 and before the patch:

STM32MP> env erase

Erasing Environment on MMC...

Caution! Your devices Erase group is 0x400 The erase range would be change to 0x2000~0x27ff

16 blocks erased: OK

Caution! Your devices Erase group is 0x400 The erase range would be change to 0x2000~0x23ff

16 blocks erased: OK OK

After this patch: STM32MP> env erase Erasing Environment on MMC... 1024 blocks erased at 0x2000: OK 1024 blocks erased at 0x2000: OK OK

Here the 2 copies of U-Boot environment are in the same devices Erase group: it is erased twice.

Signed-off-by: Patrick Delaunay patrick.delaunay@foss.st.com ---

env/mmc.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/env/mmc.c b/env/mmc.c index 465b104559..0c498d9a46 100644 --- a/env/mmc.c +++ b/env/mmc.c @@ -257,12 +257,15 @@ static inline int erase_env(struct mmc *mmc, unsigned long size, { uint blk_start, blk_cnt, n; struct blk_desc *desc = mmc_get_blk_desc(mmc); + u32 erase_size;

- blk_start = ALIGN(offset, mmc->write_bl_len) / mmc->write_bl_len; - blk_cnt = ALIGN(size, mmc->write_bl_len) / mmc->write_bl_len; + erase_size = mmc->erase_grp_size * desc->blksz; + blk_start = ALIGN_DOWN(offset, erase_size) / desc->blksz; + blk_cnt = ALIGN(size, erase_size) / desc->blksz;

n = blk_derase(desc, blk_start, blk_cnt); - printf("%d blocks erased: %s\n", n, (n == blk_cnt) ? "OK" : "ERROR"); + printf("%d blocks erased at 0x%x: %s\n", n, blk_start, + (n == blk_cnt) ? "OK" : "ERROR");

return (n == blk_cnt) ? 0 : 1; } @@ -286,6 +289,7 @@ static int env_mmc_erase(void) goto fini; }

+ printf("\n"); ret = erase_env(mmc, CONFIG_ENV_SIZE, offset);

#ifdef CONFIG_ENV_OFFSET_REDUND

On 2/16/22 00:23, Patrick Delaunay wrote:

Fix the end address in the message for unaligned erase request in mmc_berase() when start + blkcnt is aligned to erase_grp_size.

for example:

• start = 0x2000 - 26
• count = 26
• erase_grp_size = 0x400

Caution! Your devices Erase group is 0x400 The erase range would be change to 0x2000~0x27ff

But no issue when the end address is not aligned, for example

• start = 0x2000 - 2 * 26
• count = 26
• erase_grp_size = 0x400

Caution! Your devices Erase group is 0x400 The erase range would be change to 0x2000~0x23ff

Signed-off-by: Patrick Delaunay patrick.delaunay@foss.st.com

Reviewed-by: Jaehoon Chung jh80.chung@samsung.com

Best Regards, Jaehoon Chung

---

drivers/mmc/mmc_write.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mmc/mmc_write.c b/drivers/mmc/mmc_write.c index d23b7d9729..eab94c7b60 100644 --- a/drivers/mmc/mmc_write.c +++ b/drivers/mmc/mmc_write.c @@ -102,7 +102,7 @@ ulong mmc_berase(struct blk_desc *block_dev, lbaint_t start, lbaint_t blkcnt) "The erase range would be change to " "0x" LBAF "~0x" LBAF "\n\n", mmc->erase_grp_size, start & ~(mmc->erase_grp_size - 1),

•        ((start + blkcnt + mmc->erase_grp_size)
  

•        ((start + blkcnt + mmc->erase_grp_size - 1)
    & ~(mmc->erase_grp_size - 1)) - 1);
  
  

  while (blk < blkcnt) {