* [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data
@ 2025-06-05 1:07 Masami Hiramatsu (Google)
2025-06-05 1:14 ` Sergey Senozhatsky
` (2 more replies)
0 siblings, 3 replies; 5+ messages in thread
From: Masami Hiramatsu (Google) @ 2025-06-05 1:07 UTC (permalink / raw)
To: Chaotian Jing, Ulf Hansson, Matthias Brugger,
AngeloGioacchino Del Regno
Cc: mhiramat, Sergey Senozhatsky, linux-mmc, linux-kernel,
linux-arm-kernel, linux-mediatek
From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
mtk-msdc driver causes a kernel crash after swiotlb buffer is full.
---
mtk-msdc 11240000.mmc: swiotlb buffer is full (sz: 16384 bytes), total 32768 (slots), used 32732 (slots)
mtk-msdc 11240000.mmc: msdc_track_cmd_data: cmd=18 arg=0397A6F8; host->error=0x00000004
Unable to handle kernel paging request at virtual address ffffffffc0001fc0
---
When swiotlb buffer is full, the dma_map_sg() returns 0 to
msdc_prepare_data(), but it does not check it and sets the
MSDC_PREPARE_FLAG.
swiotlb_tbl_map_single() /* prints "swiotlb buffer is full" */
<-swiotlb_map()
<-dma_direct_map_page()
<-dma_direct_map_sg()
<-__dma_map_sg_attrs()
<-dma_map_sg_attrs()
<-dma_map_sg() /* returns 0 (pages mapped) */
<-msdc_prepare_data()
Then, the msdc_unprepare_data() checks MSDC_PREPARE_FLAG and calls
dma_unmap_sg() with unmapped pages. It causes a page fault.
To fix this problem, Do not set MSDC_PREPARE_FLAG if dma_map_sg()
fails because this is not prepared.
Fixes: 208489032bdd ("mmc: mediatek: Add Mediatek MMC driver")
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
---
drivers/mmc/host/mtk-sd.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
index 345ea91629e0..3594010bc229 100644
--- a/drivers/mmc/host/mtk-sd.c
+++ b/drivers/mmc/host/mtk-sd.c
@@ -827,9 +827,10 @@ static inline void msdc_dma_setup(struct msdc_host *host, struct msdc_dma *dma,
static void msdc_prepare_data(struct msdc_host *host, struct mmc_data *data)
{
if (!(data->host_cookie & MSDC_PREPARE_FLAG)) {
- data->host_cookie |= MSDC_PREPARE_FLAG;
data->sg_count = dma_map_sg(host->dev, data->sg, data->sg_len,
mmc_get_dma_dir(data));
+ if (data->sg_count)
+ data->host_cookie |= MSDC_PREPARE_FLAG;
}
}
^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data
2025-06-05 1:07 [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data Masami Hiramatsu (Google)
@ 2025-06-05 1:14 ` Sergey Senozhatsky
2025-06-05 5:56 ` AngeloGioacchino Del Regno
2025-06-09 14:25 ` Ulf Hansson
2 siblings, 0 replies; 5+ messages in thread
From: Sergey Senozhatsky @ 2025-06-05 1:14 UTC (permalink / raw)
To: Masami Hiramatsu (Google)
Cc: Chaotian Jing, Ulf Hansson, Matthias Brugger,
AngeloGioacchino Del Regno, Sergey Senozhatsky, linux-mmc,
linux-kernel, linux-arm-kernel, linux-mediatek
On (25/06/05 10:07), Masami Hiramatsu (Google) wrote:
> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>
> mtk-msdc driver causes a kernel crash after swiotlb buffer is full.
>
> ---
> mtk-msdc 11240000.mmc: swiotlb buffer is full (sz: 16384 bytes), total 32768 (slots), used 32732 (slots)
> mtk-msdc 11240000.mmc: msdc_track_cmd_data: cmd=18 arg=0397A6F8; host->error=0x00000004
> Unable to handle kernel paging request at virtual address ffffffffc0001fc0
> ---
>
> When swiotlb buffer is full, the dma_map_sg() returns 0 to
> msdc_prepare_data(), but it does not check it and sets the
> MSDC_PREPARE_FLAG.
>
> swiotlb_tbl_map_single() /* prints "swiotlb buffer is full" */
> <-swiotlb_map()
> <-dma_direct_map_page()
> <-dma_direct_map_sg()
> <-__dma_map_sg_attrs()
> <-dma_map_sg_attrs()
> <-dma_map_sg() /* returns 0 (pages mapped) */
> <-msdc_prepare_data()
>
> Then, the msdc_unprepare_data() checks MSDC_PREPARE_FLAG and calls
> dma_unmap_sg() with unmapped pages. It causes a page fault.
>
> To fix this problem, Do not set MSDC_PREPARE_FLAG if dma_map_sg()
> fails because this is not prepared.
>
> Fixes: 208489032bdd ("mmc: mediatek: Add Mediatek MMC driver")
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Tested-by: Sergey Senozhatsky <senozhatsky@chromium.org>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data
2025-06-05 1:07 [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data Masami Hiramatsu (Google)
2025-06-05 1:14 ` Sergey Senozhatsky
@ 2025-06-05 5:56 ` AngeloGioacchino Del Regno
2025-06-09 14:25 ` Ulf Hansson
2 siblings, 0 replies; 5+ messages in thread
From: AngeloGioacchino Del Regno @ 2025-06-05 5:56 UTC (permalink / raw)
To: Masami Hiramatsu (Google), Chaotian Jing, Ulf Hansson,
Matthias Brugger
Cc: Sergey Senozhatsky, linux-mmc, linux-kernel, linux-arm-kernel,
linux-mediatek
Il 05/06/25 03:07, Masami Hiramatsu (Google) ha scritto:
> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>
> mtk-msdc driver causes a kernel crash after swiotlb buffer is full.
>
> ---
> mtk-msdc 11240000.mmc: swiotlb buffer is full (sz: 16384 bytes), total 32768 (slots), used 32732 (slots)
> mtk-msdc 11240000.mmc: msdc_track_cmd_data: cmd=18 arg=0397A6F8; host->error=0x00000004
> Unable to handle kernel paging request at virtual address ffffffffc0001fc0
> ---
>
> When swiotlb buffer is full, the dma_map_sg() returns 0 to
> msdc_prepare_data(), but it does not check it and sets the
> MSDC_PREPARE_FLAG.
>
> swiotlb_tbl_map_single() /* prints "swiotlb buffer is full" */
> <-swiotlb_map()
> <-dma_direct_map_page()
> <-dma_direct_map_sg()
> <-__dma_map_sg_attrs()
> <-dma_map_sg_attrs()
> <-dma_map_sg() /* returns 0 (pages mapped) */
> <-msdc_prepare_data()
>
> Then, the msdc_unprepare_data() checks MSDC_PREPARE_FLAG and calls
> dma_unmap_sg() with unmapped pages. It causes a page fault.
>
> To fix this problem, Do not set MSDC_PREPARE_FLAG if dma_map_sg()
> fails because this is not prepared.
>
> Fixes: 208489032bdd ("mmc: mediatek: Add Mediatek MMC driver")
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data
2025-06-05 1:07 [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data Masami Hiramatsu (Google)
2025-06-05 1:14 ` Sergey Senozhatsky
2025-06-05 5:56 ` AngeloGioacchino Del Regno
@ 2025-06-09 14:25 ` Ulf Hansson
2025-06-10 4:41 ` Masami Hiramatsu
2 siblings, 1 reply; 5+ messages in thread
From: Ulf Hansson @ 2025-06-09 14:25 UTC (permalink / raw)
To: Masami Hiramatsu (Google)
Cc: Chaotian Jing, Matthias Brugger, AngeloGioacchino Del Regno,
Sergey Senozhatsky, linux-mmc, linux-kernel, linux-arm-kernel,
linux-mediatek
On Thu, 5 Jun 2025 at 03:07, Masami Hiramatsu (Google)
<mhiramat@kernel.org> wrote:
>
> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>
> mtk-msdc driver causes a kernel crash after swiotlb buffer is full.
>
> ---
> mtk-msdc 11240000.mmc: swiotlb buffer is full (sz: 16384 bytes), total 32768 (slots), used 32732 (slots)
> mtk-msdc 11240000.mmc: msdc_track_cmd_data: cmd=18 arg=0397A6F8; host->error=0x00000004
> Unable to handle kernel paging request at virtual address ffffffffc0001fc0
> ---
>
> When swiotlb buffer is full, the dma_map_sg() returns 0 to
> msdc_prepare_data(), but it does not check it and sets the
> MSDC_PREPARE_FLAG.
>
> swiotlb_tbl_map_single() /* prints "swiotlb buffer is full" */
> <-swiotlb_map()
> <-dma_direct_map_page()
> <-dma_direct_map_sg()
> <-__dma_map_sg_attrs()
> <-dma_map_sg_attrs()
> <-dma_map_sg() /* returns 0 (pages mapped) */
> <-msdc_prepare_data()
>
> Then, the msdc_unprepare_data() checks MSDC_PREPARE_FLAG and calls
> dma_unmap_sg() with unmapped pages. It causes a page fault.
>
> To fix this problem, Do not set MSDC_PREPARE_FLAG if dma_map_sg()
> fails because this is not prepared.
>
> Fixes: 208489032bdd ("mmc: mediatek: Add Mediatek MMC driver")
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Applied for fixes and by adding a stable-tag, thanks!
Note that it looked like the patch was not entirely correctly
formatted, but I fixed it up when applying. Please have a look at the
fixes branch to make sure things look okay to you.
Kind regards
Uffe
> ---
> drivers/mmc/host/mtk-sd.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
> index 345ea91629e0..3594010bc229 100644
> --- a/drivers/mmc/host/mtk-sd.c
> +++ b/drivers/mmc/host/mtk-sd.c
> @@ -827,9 +827,10 @@ static inline void msdc_dma_setup(struct msdc_host *host, struct msdc_dma *dma,
> static void msdc_prepare_data(struct msdc_host *host, struct mmc_data *data)
> {
> if (!(data->host_cookie & MSDC_PREPARE_FLAG)) {
> - data->host_cookie |= MSDC_PREPARE_FLAG;
> data->sg_count = dma_map_sg(host->dev, data->sg, data->sg_len,
> mmc_get_dma_dir(data));
> + if (data->sg_count)
> + data->host_cookie |= MSDC_PREPARE_FLAG;
> }
> }
>
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data
2025-06-09 14:25 ` Ulf Hansson
@ 2025-06-10 4:41 ` Masami Hiramatsu
0 siblings, 0 replies; 5+ messages in thread
From: Masami Hiramatsu @ 2025-06-10 4:41 UTC (permalink / raw)
To: Ulf Hansson
Cc: Chaotian Jing, Matthias Brugger, AngeloGioacchino Del Regno,
Sergey Senozhatsky, linux-mmc, linux-kernel, linux-arm-kernel,
linux-mediatek
On Mon, 9 Jun 2025 16:25:24 +0200
Ulf Hansson <ulf.hansson@linaro.org> wrote:
> On Thu, 5 Jun 2025 at 03:07, Masami Hiramatsu (Google)
> <mhiramat@kernel.org> wrote:
> >
> > From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
> >
> > mtk-msdc driver causes a kernel crash after swiotlb buffer is full.
> >
> > ---
> > mtk-msdc 11240000.mmc: swiotlb buffer is full (sz: 16384 bytes), total 32768 (slots), used 32732 (slots)
> > mtk-msdc 11240000.mmc: msdc_track_cmd_data: cmd=18 arg=0397A6F8; host->error=0x00000004
> > Unable to handle kernel paging request at virtual address ffffffffc0001fc0
> > ---
> >
> > When swiotlb buffer is full, the dma_map_sg() returns 0 to
> > msdc_prepare_data(), but it does not check it and sets the
> > MSDC_PREPARE_FLAG.
> >
> > swiotlb_tbl_map_single() /* prints "swiotlb buffer is full" */
> > <-swiotlb_map()
> > <-dma_direct_map_page()
> > <-dma_direct_map_sg()
> > <-__dma_map_sg_attrs()
> > <-dma_map_sg_attrs()
> > <-dma_map_sg() /* returns 0 (pages mapped) */
> > <-msdc_prepare_data()
> >
> > Then, the msdc_unprepare_data() checks MSDC_PREPARE_FLAG and calls
> > dma_unmap_sg() with unmapped pages. It causes a page fault.
> >
> > To fix this problem, Do not set MSDC_PREPARE_FLAG if dma_map_sg()
> > fails because this is not prepared.
> >
> > Fixes: 208489032bdd ("mmc: mediatek: Add Mediatek MMC driver")
> > Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>
> Applied for fixes and by adding a stable-tag, thanks!
>
> Note that it looked like the patch was not entirely correctly
> formatted, but I fixed it up when applying. Please have a look at the
> fixes branch to make sure things look okay to you.
Thank for applying and fixing!
Thank you,
>
> Kind regards
> Uffe
>
>
>
> > ---
> > drivers/mmc/host/mtk-sd.c | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
> > index 345ea91629e0..3594010bc229 100644
> > --- a/drivers/mmc/host/mtk-sd.c
> > +++ b/drivers/mmc/host/mtk-sd.c
> > @@ -827,9 +827,10 @@ static inline void msdc_dma_setup(struct msdc_host *host, struct msdc_dma *dma,
> > static void msdc_prepare_data(struct msdc_host *host, struct mmc_data *data)
> > {
> > if (!(data->host_cookie & MSDC_PREPARE_FLAG)) {
> > - data->host_cookie |= MSDC_PREPARE_FLAG;
> > data->sg_count = dma_map_sg(host->dev, data->sg, data->sg_len,
> > mmc_get_dma_dir(data));
> > + if (data->sg_count)
> > + data->host_cookie |= MSDC_PREPARE_FLAG;
> > }
> > }
> >
> >
--
Masami Hiramatsu (Google) <mhiramat@kernel.org>
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2025-06-10 4:43 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-06-05 1:07 [PATCH] mtk-sd: Fix a pagefault in dma_unmap_sg() for not prepared data Masami Hiramatsu (Google)
2025-06-05 1:14 ` Sergey Senozhatsky
2025-06-05 5:56 ` AngeloGioacchino Del Regno
2025-06-09 14:25 ` Ulf Hansson
2025-06-10 4:41 ` Masami Hiramatsu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).