mmc: fix the mmc_app_send_scr() for dma transfer
PATCH RESEND 1/1 mmc: fix the mmc_app_send_scr() for dma transfer
This patch is based on the commit "af51715079e7fb6b290e1881d63d815dc4de5011":
* Bugfix to that mmc_send_cxd_data() code: dma-to-stack is unsafe/nonportable, so kmalloc a bounce buffer instead.
The driver may invalidate the mmc_card->csd when host driver uses dma. So the subroutine also needs kmalloc buffer.
--- drivers/mmc/core/sd_ops.c | 14 +++++++++++++- 1 files changed, 13 insertions(+), 1 deletions(-)
drivers/mmc/core/sd_ops.c +13 -1
@@ -9,6 +9,7 @@* your option) any later version. */ +#include <linux/slab.h> #include <linux/types.h> #include <linux/scatterlist.h>
@@ -252,6 +253,7 @@ int mmc_app_send_scr(struct mmc_card *card, u32 *scr)struct mmc_command cmd;struct mmc_data data;struct scatterlist sg; +void *data_buf;BUG_ON (!card);BUG_ON (!card->host);
@@ -263,6 +265,13 @@ int mmc_app_send_scr(struct mmc_card *card, u32 *scr)if (err)return err; +/* dma onto stack is unsafe/nonportable, but callers to this +* routine normally provide temporary on-stack buffers ... +*/ + data_buf =kmalloc (sizeof (card->raw_scr),GFP_KERNEL ); +if (data_buf ==NULL ) +return -ENOMEM ; +memset (&mrq,0 ,sizeof (struct mmc_request ));memset (&cmd,0 ,sizeof (struct mmc_command ));memset (&data,0 ,sizeof (struct mmc_data ));
@@ -280,12 +289,15 @@ int mmc_app_send_scr(struct mmc_card *card, u32 *scr) data.sg = &sg; data.sg_len =1 ; -sg_init_one (&sg, scr,8 ); +sg_init_one (&sg, data_buf,8 );mmc_set_data_timeout (&data, card);mmc_wait_for_req (card->host, &mrq); +memcpy (scr, data_buf,sizeof (card->raw_scr)); +kfree (data_buf); +if (cmd.error)return cmd.error;if (data.error)
Signature · 1 line
1.7.1
Hi Yoshihiro,
Quoted 9 lines · Yoshihiro Shimoda
This patch is based on the commit "
af51715079e7fb6b290e1881d63d815dc4de5011":* Bugfix to that
mmc_send_cxd_data()code: dma-to-stack is unsafe/nonportable, so kmalloc a bounce buffer instead.The driver may invalidate the
mmc_card->csdwhen host driver uses dma. So the subroutine also needs kmalloc buffer.Signed-off-by Yoshihiro Shimoda <[email protected]>
Thanks for the resend, pushed to mmc-next for .39-rc2.
- Chris.
Signature · 2 lines
Chris Ball <[email protected]> <http://printf.net/> One Laptop Per Child