This patch series fixes warnings see with DMA_API_DEBUG_SG=y
Ludovic Barre (2):
mmc: mmci_sdmmc: fix DMA API warning overlapping mappings
mmc: mmci_sdmmc: fix DMA API warning max segment size
drivers/mmc/host/mmci_stm32_sdmmc.c | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
--
2.17.1
Turning on CONFIG_DMA_API_DEBUG_SG results in the following warning:
WARNING: CPU: 1 PID: 85 at kernel/dma/debug.c:1302 debug_dma_map_sg+0x2a0/0x3cc
mmci-pl18x 58005000.sdmmc: DMA-API: mapping sg segment longer than device claims to support [len=126976] [max=65536]
dma api debug checks and compares the segment size to
dma_get_max_seg_size (dev->dma_parms->max_segment_size),
the sdmmc variant has an internal DMA and should define
its max_segment_size constraint to avoid this warning.
This Patch defines the dev->dma_parms->max_segment_size
with the constraint already set for mmc core
(host->mmc->max_seg_size).
Signed-off-by: Ludovic Barre <[email protected]>
---
drivers/mmc/host/mmci_stm32_sdmmc.c | 11 +++++------
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
index 2965b1c062e1..51db30acf4dc 100644
--- a/drivers/mmc/host/mmci_stm32_sdmmc.c
+++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
@@ -119,20 +119,19 @@ static void sdmmc_idma_unprep_data(struct mmci_host *host,
static int sdmmc_idma_setup(struct mmci_host *host)
{
struct sdmmc_idma *idma;
+ struct device *dev = mmc_dev(host->mmc);
- idma = devm_kzalloc(mmc_dev(host->mmc), sizeof(*idma), GFP_KERNEL);
+ idma = devm_kzalloc(dev, sizeof(*idma), GFP_KERNEL);
if (!idma)
return -ENOMEM;
host->dma_priv = idma;
if (host->variant->dma_lli) {
- idma->sg_cpu = dmam_alloc_coherent(mmc_dev(host->mmc),
- SDMMC_LLI_BUF_LEN,
+ idma->sg_cpu = dmam_alloc_coherent(dev, SDMMC_LLI_BUF_LEN,
&idma->sg_dma, GFP_KERNEL);
if (!idma->sg_cpu) {
- dev_err(mmc_dev(host->mmc),
- "Failed to alloc IDMA descriptor\n");
+ dev_err(dev, "Failed to alloc IDMA descriptor\n");
return -ENOMEM;
}
host->mmc->max_segs = SDMMC_LLI_BUF_LEN /
@@ -143,7 +142,7 @@ static int sdmmc_idma_setup(struct mmci_host *host)
host->mmc->max_seg_size = host->mmc->max_req_size;
}
- return 0;
+ return dma_set_max_seg_size(dev, host->mmc->max_seg_size);
}
static int sdmmc_idma_start(struct mmci_host *host, unsigned int *datactrl)
--
2.17.1
Turning on CONFIG_DMA_API_DEBUG_SG results in the following warning:
WARNING: CPU: 1 PID: 20 at kernel/dma/debug.c:500 add_dma_entry+0x16c/0x17c
DMA-API: exceeded 7 overlapping mappings of cacheline 0x031d2645
Modules linked in:
CPU: 1 PID: 20 Comm: kworker/1:1 Not tainted 5.5.0-rc2-00021-gdeda30999c2b-dirty #49
Hardware name: STM32 (Device Tree Support)
Workqueue: events_freezable mmc_rescan
[<c03138c0>] (unwind_backtrace) from [<c030d760>] (show_stack+0x10/0x14)
[<c030d760>] (show_stack) from [<c0f2eb28>] (dump_stack+0xc0/0xd4)
[<c0f2eb28>] (dump_stack) from [<c034a14c>] (__warn+0xd0/0xf8)
[<c034a14c>] (__warn) from [<c034a530>] (warn_slowpath_fmt+0x94/0xb8)
[<c034a530>] (warn_slowpath_fmt) from [<c03bca0c>] (add_dma_entry+0x16c/0x17c)
[<c03bca0c>] (add_dma_entry) from [<c03bdf54>] (debug_dma_map_sg+0xe4/0x3d4)
[<c03bdf54>] (debug_dma_map_sg) from [<c0d09244>] (sdmmc_idma_prep_data+0x94/0xf8)
[<c0d09244>] (sdmmc_idma_prep_data) from [<c0d05a2c>] (mmci_prep_data+0x2c/0xb0)
[<c0d05a2c>] (mmci_prep_data) from [<c0d073ec>] (mmci_start_data+0x134/0x2f0)
[<c0d073ec>] (mmci_start_data) from [<c0d078d0>] (mmci_request+0xe8/0x154)
[<c0d078d0>] (mmci_request) from [<c0cecb44>] (mmc_start_request+0x94/0xbc)
DMA api debug brings to light leaking dma-mappings, dma_map_sg and
dma_unmap_sg are not correctly balanced.
If a request is prepared, the dma_map/unmap are done in asynchronous
call pre_req (prep_data) and post_req (unprep_data). In this case
the dma-mapping is right balanced.
But if the request was not prepared, the data->host_cookie is
define to zero and the dma_map/unmap must be done in the request.
The dma_map is called by mmci_dma_start (prep_data), but there is
no dma_unmap in this case.
This patch adds dma_unmap_sg when the dma is finalized and
the data cookie is zero (request not prepared).
Signed-off-by: Ludovic Barre <[email protected]>
---
drivers/mmc/host/mmci_stm32_sdmmc.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
index 14f99d8aa3f0..2965b1c062e1 100644
--- a/drivers/mmc/host/mmci_stm32_sdmmc.c
+++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
@@ -188,6 +188,9 @@ static int sdmmc_idma_start(struct mmci_host *host, unsigned int *datactrl)
static void sdmmc_idma_finalize(struct mmci_host *host, struct mmc_data *data)
{
writel_relaxed(0, host->base + MMCI_STM32_IDMACTRLR);
+
+ if (!data->host_cookie)
+ sdmmc_idma_unprep_data(host, data, 0);
}
static void mmci_sdmmc_set_clkreg(struct mmci_host *host, unsigned int desired)
--
2.17.1
On Tue, 26 May 2020 at 17:51, Ludovic Barre <[email protected]> wrote:
>
> Turning on CONFIG_DMA_API_DEBUG_SG results in the following warning:
> WARNING: CPU: 1 PID: 85 at kernel/dma/debug.c:1302 debug_dma_map_sg+0x2a0/0x3cc
> mmci-pl18x 58005000.sdmmc: DMA-API: mapping sg segment longer than device claims to support [len=126976] [max=65536]
>
> dma api debug checks and compares the segment size to
> dma_get_max_seg_size (dev->dma_parms->max_segment_size),
> the sdmmc variant has an internal DMA and should define
> its max_segment_size constraint to avoid this warning.
>
> This Patch defines the dev->dma_parms->max_segment_size
> with the constraint already set for mmc core
> (host->mmc->max_seg_size).
>
> Signed-off-by: Ludovic Barre <[email protected]>
Applied for next, thanks!
Note, a manual backport is needed for stable, as
dma_set_max_seg_size() will fail for older kernels.
We needed to revert 9495b7e92f7 ("driver core: platform: Initialize
dma_parms for platform devices"), for stable kernels [1].
Kind regards
Uffe
[1]
https://lkml.org/lkml/2020/5/26/1216
> ---
> drivers/mmc/host/mmci_stm32_sdmmc.c | 11 +++++------
> 1 file changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
> index 2965b1c062e1..51db30acf4dc 100644
> --- a/drivers/mmc/host/mmci_stm32_sdmmc.c
> +++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
> @@ -119,20 +119,19 @@ static void sdmmc_idma_unprep_data(struct mmci_host *host,
> static int sdmmc_idma_setup(struct mmci_host *host)
> {
> struct sdmmc_idma *idma;
> + struct device *dev = mmc_dev(host->mmc);
>
> - idma = devm_kzalloc(mmc_dev(host->mmc), sizeof(*idma), GFP_KERNEL);
> + idma = devm_kzalloc(dev, sizeof(*idma), GFP_KERNEL);
> if (!idma)
> return -ENOMEM;
>
> host->dma_priv = idma;
>
> if (host->variant->dma_lli) {
> - idma->sg_cpu = dmam_alloc_coherent(mmc_dev(host->mmc),
> - SDMMC_LLI_BUF_LEN,
> + idma->sg_cpu = dmam_alloc_coherent(dev, SDMMC_LLI_BUF_LEN,
> &idma->sg_dma, GFP_KERNEL);
> if (!idma->sg_cpu) {
> - dev_err(mmc_dev(host->mmc),
> - "Failed to alloc IDMA descriptor\n");
> + dev_err(dev, "Failed to alloc IDMA descriptor\n");
> return -ENOMEM;
> }
> host->mmc->max_segs = SDMMC_LLI_BUF_LEN /
> @@ -143,7 +142,7 @@ static int sdmmc_idma_setup(struct mmci_host *host)
> host->mmc->max_seg_size = host->mmc->max_req_size;
> }
>
> - return 0;
> + return dma_set_max_seg_size(dev, host->mmc->max_seg_size);
> }
>
> static int sdmmc_idma_start(struct mmci_host *host, unsigned int *datactrl)
> --
> 2.17.1
>
On Tue, 26 May 2020 at 17:51, Ludovic Barre <[email protected]> wrote:
>
> Turning on CONFIG_DMA_API_DEBUG_SG results in the following warning:
> WARNING: CPU: 1 PID: 20 at kernel/dma/debug.c:500 add_dma_entry+0x16c/0x17c
> DMA-API: exceeded 7 overlapping mappings of cacheline 0x031d2645
> Modules linked in:
> CPU: 1 PID: 20 Comm: kworker/1:1 Not tainted 5.5.0-rc2-00021-gdeda30999c2b-dirty #49
> Hardware name: STM32 (Device Tree Support)
> Workqueue: events_freezable mmc_rescan
> [<c03138c0>] (unwind_backtrace) from [<c030d760>] (show_stack+0x10/0x14)
> [<c030d760>] (show_stack) from [<c0f2eb28>] (dump_stack+0xc0/0xd4)
> [<c0f2eb28>] (dump_stack) from [<c034a14c>] (__warn+0xd0/0xf8)
> [<c034a14c>] (__warn) from [<c034a530>] (warn_slowpath_fmt+0x94/0xb8)
> [<c034a530>] (warn_slowpath_fmt) from [<c03bca0c>] (add_dma_entry+0x16c/0x17c)
> [<c03bca0c>] (add_dma_entry) from [<c03bdf54>] (debug_dma_map_sg+0xe4/0x3d4)
> [<c03bdf54>] (debug_dma_map_sg) from [<c0d09244>] (sdmmc_idma_prep_data+0x94/0xf8)
> [<c0d09244>] (sdmmc_idma_prep_data) from [<c0d05a2c>] (mmci_prep_data+0x2c/0xb0)
> [<c0d05a2c>] (mmci_prep_data) from [<c0d073ec>] (mmci_start_data+0x134/0x2f0)
> [<c0d073ec>] (mmci_start_data) from [<c0d078d0>] (mmci_request+0xe8/0x154)
> [<c0d078d0>] (mmci_request) from [<c0cecb44>] (mmc_start_request+0x94/0xbc)
>
> DMA api debug brings to light leaking dma-mappings, dma_map_sg and
> dma_unmap_sg are not correctly balanced.
>
> If a request is prepared, the dma_map/unmap are done in asynchronous
> call pre_req (prep_data) and post_req (unprep_data). In this case
> the dma-mapping is right balanced.
>
> But if the request was not prepared, the data->host_cookie is
> define to zero and the dma_map/unmap must be done in the request.
> The dma_map is called by mmci_dma_start (prep_data), but there is
> no dma_unmap in this case.
>
> This patch adds dma_unmap_sg when the dma is finalized and
> the data cookie is zero (request not prepared).
>
> Signed-off-by: Ludovic Barre <[email protected]>
Applied for next by adding a fixes tag and a stable tag, thanks!
Fixes: 46b723dd867d ("mmc: mmci: add stm32 sdmmc variant")
Kind regards
Uffe
> ---
> drivers/mmc/host/mmci_stm32_sdmmc.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
> index 14f99d8aa3f0..2965b1c062e1 100644
> --- a/drivers/mmc/host/mmci_stm32_sdmmc.c
> +++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
> @@ -188,6 +188,9 @@ static int sdmmc_idma_start(struct mmci_host *host, unsigned int *datactrl)
> static void sdmmc_idma_finalize(struct mmci_host *host, struct mmc_data *data)
> {
> writel_relaxed(0, host->base + MMCI_STM32_IDMACTRLR);
> +
> + if (!data->host_cookie)
> + sdmmc_idma_unprep_data(host, data, 0);
> }
>
> static void mmci_sdmmc_set_clkreg(struct mmci_host *host, unsigned int desired)
> --
> 2.17.1
>