It's currently the amba/platform driver's responsibility to initialize the
pointer, dma_parms, for its corresponding struct device. The benefit with this
approach allows us to avoid the initialization and to not waste memory for the
struct device_dma_parameters, as this can be decided on a case by case basis.
However, it has turned out that this approach is not very practical. Not only
does it lead to open coding, but also to real errors. In principle callers of
dma_set_max_seg_size() doesn't check the error code, but just assumes it
succeeds.
For these reasons, this series initializes the dma_parms from the amba/platform
bus at the device registration point. This also follows the way the PCI devices
are being managed, see pci_device_add().
If it turns out that this is an acceptable solution, we probably also want the
changes for stable, but I am not sure if it applies without conflicts.
The series is based on v5.6-rc7.
Kind regards
Ulf Hansson
Ulf Hansson (2):
driver core: platform: Initialize dma_parms for platform devices
amba: Initialize dma_parms for amba devices
drivers/amba/bus.c | 2 ++
drivers/base/platform.c | 1 +
include/linux/amba/bus.h | 1 +
include/linux/platform_device.h | 1 +
4 files changed, 5 insertions(+)
--
2.20.1
It's currently the platform driver's responsibility to initialize the
pointer, dma_parms, for its corresponding struct device. The benefit with
this approach allows us to avoid the initialization and to not waste memory
for the struct device_dma_parameters, as this can be decided on a case by
case basis.
However, it has turned out that this approach is not very practical. Not
only does it lead to open coding, but also to real errors. In principle
callers of dma_set_max_seg_size() doesn't check the error code, but just
assumes it succeeds.
For these reasons, let's do the initialization from the common platform bus
at the device registration point. This also follows the way the PCI devices
are being managed, see pci_device_add().
Suggested-by: Christoph Hellwig <[email protected]>
Cc: <[email protected]>
Signed-off-by: Ulf Hansson <[email protected]>
---
drivers/base/platform.c | 1 +
include/linux/platform_device.h | 1 +
2 files changed, 2 insertions(+)
diff --git a/drivers/base/platform.c b/drivers/base/platform.c
index b5ce7b085795..46abbfb52655 100644
--- a/drivers/base/platform.c
+++ b/drivers/base/platform.c
@@ -512,6 +512,7 @@ int platform_device_add(struct platform_device *pdev)
pdev->dev.parent = &platform_bus;
pdev->dev.bus = &platform_bus_type;
+ pdev->dev.dma_parms = &pdev->dma_parms;
switch (pdev->id) {
default:
diff --git a/include/linux/platform_device.h b/include/linux/platform_device.h
index 041bfa412aa0..81900b3cbe37 100644
--- a/include/linux/platform_device.h
+++ b/include/linux/platform_device.h
@@ -25,6 +25,7 @@ struct platform_device {
bool id_auto;
struct device dev;
u64 platform_dma_mask;
+ struct device_dma_parameters dma_parms;
u32 num_resources;
struct resource *resource;
--
2.20.1
It's currently the amba driver's responsibility to initialize the pointer,
dma_parms, for its corresponding struct device. The benefit with this
approach allows us to avoid the initialization and to not waste memory for
the struct device_dma_parameters, as this can be decided on a case by case
basis.
However, it has turned out that this approach is not very practical. Not
only does it lead to open coding, but also to real errors. In principle
callers of dma_set_max_seg_size() doesn't check the error code, but just
assumes it succeeds.
For these reasons, let's do the initialization from the common amba bus at
the device registration point. This also follows the way the PCI devices
are being managed, see pci_device_add().
Suggested-by: Christoph Hellwig <[email protected]>
Cc: Russell King <[email protected]>
Cc: <[email protected]>
Signed-off-by: Ulf Hansson <[email protected]>
---
drivers/amba/bus.c | 2 ++
include/linux/amba/bus.h | 1 +
2 files changed, 3 insertions(+)
diff --git a/drivers/amba/bus.c b/drivers/amba/bus.c
index fe1523664816..5e61783ce92d 100644
--- a/drivers/amba/bus.c
+++ b/drivers/amba/bus.c
@@ -374,6 +374,8 @@ static int amba_device_try_add(struct amba_device *dev, struct resource *parent)
WARN_ON(dev->irq[0] == (unsigned int)-1);
WARN_ON(dev->irq[1] == (unsigned int)-1);
+ dev->dev.dma_parms = &dev->dma_parms;
+
ret = request_resource(parent, &dev->res);
if (ret)
goto err_out;
diff --git a/include/linux/amba/bus.h b/include/linux/amba/bus.h
index 26f0ecf401ea..0bbfd647f5c6 100644
--- a/include/linux/amba/bus.h
+++ b/include/linux/amba/bus.h
@@ -65,6 +65,7 @@ struct amba_device {
struct device dev;
struct resource res;
struct clk *pclk;
+ struct device_dma_parameters dma_parms;
unsigned int periphid;
unsigned int cid;
struct amba_cs_uci_id uci;
--
2.20.1
Le 3/25/20 à 12:34 PM, Ulf Hansson a écrit :
> It's currently the platform driver's responsibility to initialize the
> pointer, dma_parms, for its corresponding struct device. The benefit with
> this approach allows us to avoid the initialization and to not waste memory
> for the struct device_dma_parameters, as this can be decided on a case by
> case basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common platform bus
> at the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
tested with mmc: mmci_sdmmc fix
Tested-by: Ludovic Barre <[email protected]>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
> ---
> drivers/base/platform.c | 1 +
> include/linux/platform_device.h | 1 +
> 2 files changed, 2 insertions(+)
>
> diff --git a/drivers/base/platform.c b/drivers/base/platform.c
> index b5ce7b085795..46abbfb52655 100644
> --- a/drivers/base/platform.c
> +++ b/drivers/base/platform.c
> @@ -512,6 +512,7 @@ int platform_device_add(struct platform_device *pdev)
> pdev->dev.parent = &platform_bus;
>
> pdev->dev.bus = &platform_bus_type;
> + pdev->dev.dma_parms = &pdev->dma_parms;
>
> switch (pdev->id) {
> default:
> diff --git a/include/linux/platform_device.h b/include/linux/platform_device.h
> index 041bfa412aa0..81900b3cbe37 100644
> --- a/include/linux/platform_device.h
> +++ b/include/linux/platform_device.h
> @@ -25,6 +25,7 @@ struct platform_device {
> bool id_auto;
> struct device dev;
> u64 platform_dma_mask;
> + struct device_dma_parameters dma_parms;
> u32 num_resources;
> struct resource *resource;
>
>
Le 3/25/20 à 12:34 PM, Ulf Hansson a écrit :
> It's currently the amba driver's responsibility to initialize the pointer,
> dma_parms, for its corresponding struct device. The benefit with this
> approach allows us to avoid the initialization and to not waste memory for
> the struct device_dma_parameters, as this can be decided on a case by case
> basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common amba bus at
> the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
>
tested with mmc: mmci_sdmmc fix
Tested-by: Ludovic Barre <[email protected]>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: Russell King <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
> ---
> drivers/amba/bus.c | 2 ++
> include/linux/amba/bus.h | 1 +
> 2 files changed, 3 insertions(+)
>
> diff --git a/drivers/amba/bus.c b/drivers/amba/bus.c
> index fe1523664816..5e61783ce92d 100644
> --- a/drivers/amba/bus.c
> +++ b/drivers/amba/bus.c
> @@ -374,6 +374,8 @@ static int amba_device_try_add(struct amba_device *dev, struct resource *parent)
> WARN_ON(dev->irq[0] == (unsigned int)-1);
> WARN_ON(dev->irq[1] == (unsigned int)-1);
>
> + dev->dev.dma_parms = &dev->dma_parms;
> +
> ret = request_resource(parent, &dev->res);
> if (ret)
> goto err_out;
> diff --git a/include/linux/amba/bus.h b/include/linux/amba/bus.h
> index 26f0ecf401ea..0bbfd647f5c6 100644
> --- a/include/linux/amba/bus.h
> +++ b/include/linux/amba/bus.h
> @@ -65,6 +65,7 @@ struct amba_device {
> struct device dev;
> struct resource res;
> struct clk *pclk;
> + struct device_dma_parameters dma_parms;
> unsigned int periphid;
> unsigned int cid;
> struct amba_cs_uci_id uci;
>
> -----Original Message-----
> From: Ulf Hansson <[email protected]>
> Sent: 2020??3??25?? 19:34
> To: Greg Kroah-Hartman <[email protected]>; Rafael J . Wysocki
> <[email protected]>; [email protected]
> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> Russell King <[email protected]>; Linus Walleij
> <[email protected]>; Vinod Koul <[email protected]>; BOUGH CHEN
> <[email protected]>; Ludovic Barre <[email protected]>;
> [email protected]; [email protected]; Ulf
> Hansson <[email protected]>
> Subject: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus level
>
> It's currently the amba/platform driver's responsibility to initialize the pointer,
> dma_parms, for its corresponding struct device. The benefit with this
> approach allows us to avoid the initialization and to not waste memory for
> the struct device_dma_parameters, as this can be decided on a case by case
> basis.
>
> However, it has turned out that this approach is not very practical. Not only
> does it lead to open coding, but also to real errors. In principle callers of
> dma_set_max_seg_size() doesn't check the error code, but just assumes it
> succeeds.
>
> For these reasons, this series initializes the dma_parms from the
> amba/platform bus at the device registration point. This also follows the
> way the PCI devices are being managed, see pci_device_add().
>
> If it turns out that this is an acceptable solution, we probably also want the
> changes for stable, but I am not sure if it applies without conflicts.
>
> The series is based on v5.6-rc7.
>
Hi Ulf,
Since i.MXQM SMMU related code still not upstream yet, so I apply your patches on our internal Linux branch based on v5.4.24, and find it do not work on my side. Maybe for platform core drivers, there is a gap between v5.4.24 and v5.6-rc7 which has the impact.
I will try to put our SMMU related code into v5.6-rc7, then do the test again.
Best Regards
Haibo Chen
> Kind regards
> Ulf Hansson
>
> Ulf Hansson (2):
> driver core: platform: Initialize dma_parms for platform devices
> amba: Initialize dma_parms for amba devices
>
> drivers/amba/bus.c | 2 ++
> drivers/base/platform.c | 1 +
> include/linux/amba/bus.h | 1 +
> include/linux/platform_device.h | 1 +
> 4 files changed, 5 insertions(+)
>
> --
> 2.20.1
On Wed, Mar 25, 2020 at 12:34 PM Ulf Hansson <[email protected]> wrote:
> It's currently the platform driver's responsibility to initialize the
> pointer, dma_parms, for its corresponding struct device. The benefit with
> this approach allows us to avoid the initialization and to not waste memory
> for the struct device_dma_parameters, as this can be decided on a case by
> case basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common platform bus
> at the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
This seems in line with what Christoph said.
Reviewed-by: Linus Walleij <[email protected]>
I imagine we can eventually set up more of the DMA config such
as segment size based on config from the device tree, but I'm not
sure about that yet.
Yours,
Linus Walleij
On Wed, Mar 25, 2020 at 12:34 PM Ulf Hansson <[email protected]> wrote:
> It's currently the amba driver's responsibility to initialize the pointer,
> dma_parms, for its corresponding struct device. The benefit with this
> approach allows us to avoid the initialization and to not waste memory for
> the struct device_dma_parameters, as this can be decided on a case by case
> basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common amba bus at
> the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: Russell King <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
Reviewed-by: Linus Walleij <[email protected]>
Yours,
Linus Walleij
On Wed, Mar 25, 2020 at 12:34 PM Ulf Hansson <[email protected]> wrote:
>
> It's currently the platform driver's responsibility to initialize the
> pointer, dma_parms, for its corresponding struct device. The benefit with
> this approach allows us to avoid the initialization and to not waste memory
> for the struct device_dma_parameters, as this can be decided on a case by
> case basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common platform bus
> at the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
On Wed, Mar 25, 2020 at 12:34 PM Ulf Hansson <[email protected]> wrote:
>
> It's currently the amba driver's responsibility to initialize the pointer,
> dma_parms, for its corresponding struct device. The benefit with this
> approach allows us to avoid the initialization and to not waste memory for
> the struct device_dma_parameters, as this can be decided on a case by case
> basis.
>
> However, it has turned out that this approach is not very practical. Not
> only does it lead to open coding, but also to real errors. In principle
> callers of dma_set_max_seg_size() doesn't check the error code, but just
> assumes it succeeds.
>
> For these reasons, let's do the initialization from the common amba bus at
> the device registration point. This also follows the way the PCI devices
> are being managed, see pci_device_add().
>
> Suggested-by: Christoph Hellwig <[email protected]>
> Cc: Russell King <[email protected]>
> Cc: <[email protected]>
> Signed-off-by: Ulf Hansson <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
> -----Original Message-----
> From: BOUGH CHEN
> Sent: 2020??3??26?? 12:41
> To: Ulf Hansson <[email protected]>; Greg Kroah-Hartman
> <[email protected]>; Rafael J . Wysocki <[email protected]>;
> [email protected]
> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> Russell King <[email protected]>; Linus Walleij <[email protected]>;
> Vinod Koul <[email protected]>; Ludovic Barre <[email protected]>;
> [email protected]; [email protected]
> Subject: RE: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus level
>
> > -----Original Message-----
> > From: Ulf Hansson <[email protected]>
> > Sent: 2020??3??25?? 19:34
> > To: Greg Kroah-Hartman <[email protected]>; Rafael J .
> > Wysocki <[email protected]>; [email protected]
> > Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> > Russell King <[email protected]>; Linus Walleij
> > <[email protected]>; Vinod Koul <[email protected]>; BOUGH CHEN
> > <[email protected]>; Ludovic Barre <[email protected]>;
> > [email protected]; [email protected]; Ulf
> > Hansson <[email protected]>
> > Subject: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus
> > level
> >
> > It's currently the amba/platform driver's responsibility to initialize
> > the pointer, dma_parms, for its corresponding struct device. The
> > benefit with this approach allows us to avoid the initialization and
> > to not waste memory for the struct device_dma_parameters, as this can
> > be decided on a case by case basis.
> >
> > However, it has turned out that this approach is not very practical.
> > Not only does it lead to open coding, but also to real errors. In
> > principle callers of
> > dma_set_max_seg_size() doesn't check the error code, but just assumes
> > it succeeds.
> >
> > For these reasons, this series initializes the dma_parms from the
> > amba/platform bus at the device registration point. This also follows
> > the way the PCI devices are being managed, see pci_device_add().
> >
> > If it turns out that this is an acceptable solution, we probably also
> > want the changes for stable, but I am not sure if it applies without conflicts.
> >
> > The series is based on v5.6-rc7.
> >
>
> Hi Ulf,
>
> Since i.MXQM SMMU related code still not upstream yet, so I apply your
> patches on our internal Linux branch based on v5.4.24, and find it do not work
> on my side. Maybe for platform core drivers, there is a gap between v5.4.24
> and v5.6-rc7 which has the impact.
> I will try to put our SMMU related code into v5.6-rc7, then do the test again.
>
>
Hi Ulf,
On the latest Linux-next branch, the top commit 89295c59c1f063b533d071ca49d0fa0c0783ca6f (tag: next-20200326), after add your two patches, I just add the simple debug code as following in the /driver/mmc/core/queue.c, but seems still not work as our expect, logically, it should work, so can you or anyone test on other platform? This seems weird.
diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
index 25bee3daf9e2..f091280f7ffb 100644
--- a/drivers/mmc/core/queue.c
+++ b/drivers/mmc/core/queue.c
@@ -403,6 +403,13 @@ static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
blk_queue_max_segment_size(mq->queue,
round_down(host->max_seg_size, block_size));
+ pr_err("###### the max segment size is %d\n", queue_max_segment_size(mq->queue));
+ if (host->parent->dma_parms) {
+ pr_err("######### the dma parms has value\n");
+ } else if (!(host->parent->dma_parms)) {
+ pr_err("######## the dma parms is zero!!\n");
+ }
+
dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue));
INIT_WORK(&mq->recovery_work, mmc_mq_recovery_handler);
Here is the log I got when system run, even after your patch, the dev->dma_parms is still NULL.
[ 0.989853] mmc0: new HS400 MMC card at address 0001
[ 0.995708] sdhci-esdhc-imx 30b50000.mmc: Got CD GPIO
[ 0.999374] ###### the max segment size is 65024
[ 1.008594] ######## the dma parms is zero!!
[ 1.012875] mmcblk0: mmc0:0001 IB2932 29.2 GiB
[ 1.017569] ###### the max segment size is 65024
[ 1.022195] ######## the dma parms is zero!!
[ 1.026479] mmcblk0boot0: mmc0:0001 IB2932 partition 1 4.00 MiB
[ 1.032541] ###### the max segment size is 65024
[ 1.035198] mmc1: SDHCI controller on 30b50000.mmc [30b50000.mmc] using ADMA
[ 1.037169] ######## the dma parms is zero!!
[ 1.048493] mmcblk0boot1: mmc0:0001 IB2932 partition 2 4.00 MiB
[ 1.054531] mmcblk0rpmb: mmc0:0001 IB2932 partition 3 4.00 MiB, chardev (234:0)
Regards
Haibo Chen
> Best Regards
> Haibo Chen
>
> > Kind regards
> > Ulf Hansson
> >
> > Ulf Hansson (2):
> > driver core: platform: Initialize dma_parms for platform devices
> > amba: Initialize dma_parms for amba devices
> >
> > drivers/amba/bus.c | 2 ++
> > drivers/base/platform.c | 1 +
> > include/linux/amba/bus.h | 1 +
> > include/linux/platform_device.h | 1 +
> > 4 files changed, 5 insertions(+)
> >
> > --
> > 2.20.1
On Fri, 27 Mar 2020 at 04:02, BOUGH CHEN <[email protected]> wrote:
>
>
> > -----Original Message-----
> > From: BOUGH CHEN
> > Sent: 2020年3月26日 12:41
> > To: Ulf Hansson <[email protected]>; Greg Kroah-Hartman
> > <[email protected]>; Rafael J . Wysocki <[email protected]>;
> > [email protected]
> > Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> > Russell King <[email protected]>; Linus Walleij <[email protected]>;
> > Vinod Koul <[email protected]>; Ludovic Barre <[email protected]>;
> > [email protected]; [email protected]
> > Subject: RE: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus level
> >
> > > -----Original Message-----
> > > From: Ulf Hansson <[email protected]>
> > > Sent: 2020年3月25日 19:34
> > > To: Greg Kroah-Hartman <[email protected]>; Rafael J .
> > > Wysocki <[email protected]>; [email protected]
> > > Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> > > Russell King <[email protected]>; Linus Walleij
> > > <[email protected]>; Vinod Koul <[email protected]>; BOUGH CHEN
> > > <[email protected]>; Ludovic Barre <[email protected]>;
> > > [email protected]; [email protected]; Ulf
> > > Hansson <[email protected]>
> > > Subject: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus
> > > level
> > >
> > > It's currently the amba/platform driver's responsibility to initialize
> > > the pointer, dma_parms, for its corresponding struct device. The
> > > benefit with this approach allows us to avoid the initialization and
> > > to not waste memory for the struct device_dma_parameters, as this can
> > > be decided on a case by case basis.
> > >
> > > However, it has turned out that this approach is not very practical.
> > > Not only does it lead to open coding, but also to real errors. In
> > > principle callers of
> > > dma_set_max_seg_size() doesn't check the error code, but just assumes
> > > it succeeds.
> > >
> > > For these reasons, this series initializes the dma_parms from the
> > > amba/platform bus at the device registration point. This also follows
> > > the way the PCI devices are being managed, see pci_device_add().
> > >
> > > If it turns out that this is an acceptable solution, we probably also
> > > want the changes for stable, but I am not sure if it applies without conflicts.
> > >
> > > The series is based on v5.6-rc7.
> > >
> >
> > Hi Ulf,
> >
> > Since i.MXQM SMMU related code still not upstream yet, so I apply your
> > patches on our internal Linux branch based on v5.4.24, and find it do not work
> > on my side. Maybe for platform core drivers, there is a gap between v5.4.24
> > and v5.6-rc7 which has the impact.
> > I will try to put our SMMU related code into v5.6-rc7, then do the test again.
> >
> >
>
> Hi Ulf,
>
> On the latest Linux-next branch, the top commit 89295c59c1f063b533d071ca49d0fa0c0783ca6f (tag: next-20200326), after add your two patches, I just add the simple debug code as following in the /driver/mmc/core/queue.c, but seems still not work as our expect, logically, it should work, so can you or anyone test on other platform? This seems weird.
You are right, this doesn't work for platform devices being added
through the OF path.
In other words, of_platform_device_create_pdata() manually allocates
the platform device and assigns it the &platform_bus_type, but without
calling platform_device_add().
For amba, it works fine, as in that OF path, amba_device_add() is called. Hmm.
I re-spin this, to address the problem. Perhaps we simply need to use
the ->probe() path.
Kind regards
Uffe
>
> diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
> index 25bee3daf9e2..f091280f7ffb 100644
> --- a/drivers/mmc/core/queue.c
> +++ b/drivers/mmc/core/queue.c
> @@ -403,6 +403,13 @@ static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
> blk_queue_max_segment_size(mq->queue,
> round_down(host->max_seg_size, block_size));
>
> + pr_err("###### the max segment size is %d\n", queue_max_segment_size(mq->queue));
> + if (host->parent->dma_parms) {
> + pr_err("######### the dma parms has value\n");
> + } else if (!(host->parent->dma_parms)) {
> + pr_err("######## the dma parms is zero!!\n");
> + }
> +
> dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue));
>
> INIT_WORK(&mq->recovery_work, mmc_mq_recovery_handler);
>
> Here is the log I got when system run, even after your patch, the dev->dma_parms is still NULL.
> [ 0.989853] mmc0: new HS400 MMC card at address 0001
> [ 0.995708] sdhci-esdhc-imx 30b50000.mmc: Got CD GPIO
> [ 0.999374] ###### the max segment size is 65024
> [ 1.008594] ######## the dma parms is zero!!
> [ 1.012875] mmcblk0: mmc0:0001 IB2932 29.2 GiB
> [ 1.017569] ###### the max segment size is 65024
> [ 1.022195] ######## the dma parms is zero!!
> [ 1.026479] mmcblk0boot0: mmc0:0001 IB2932 partition 1 4.00 MiB
> [ 1.032541] ###### the max segment size is 65024
> [ 1.035198] mmc1: SDHCI controller on 30b50000.mmc [30b50000.mmc] using ADMA
> [ 1.037169] ######## the dma parms is zero!!
> [ 1.048493] mmcblk0boot1: mmc0:0001 IB2932 partition 2 4.00 MiB
> [ 1.054531] mmcblk0rpmb: mmc0:0001 IB2932 partition 3 4.00 MiB, chardev (234:0)
>
>
> Regards
> Haibo Chen
> > Best Regards
> > Haibo Chen
> >
> > > Kind regards
> > > Ulf Hansson
> > >
> > > Ulf Hansson (2):
> > > driver core: platform: Initialize dma_parms for platform devices
> > > amba: Initialize dma_parms for amba devices
> > >
> > > drivers/amba/bus.c | 2 ++
> > > drivers/base/platform.c | 1 +
> > > include/linux/amba/bus.h | 1 +
> > > include/linux/platform_device.h | 1 +
> > > 4 files changed, 5 insertions(+)
> > >
> > > --
> > > 2.20.1
>
On 2020-03-27 3:34 pm, Ulf Hansson wrote:
> On Fri, 27 Mar 2020 at 04:02, BOUGH CHEN <[email protected]> wrote:
>>
>>
>>> -----Original Message-----
>>> From: BOUGH CHEN
>>> Sent: 2020年3月26日 12:41
>>> To: Ulf Hansson <[email protected]>; Greg Kroah-Hartman
>>> <[email protected]>; Rafael J . Wysocki <[email protected]>;
>>> [email protected]
>>> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
>>> Russell King <[email protected]>; Linus Walleij <[email protected]>;
>>> Vinod Koul <[email protected]>; Ludovic Barre <[email protected]>;
>>> [email protected]; [email protected]
>>> Subject: RE: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus level
>>>
>>>> -----Original Message-----
>>>> From: Ulf Hansson <[email protected]>
>>>> Sent: 2020年3月25日 19:34
>>>> To: Greg Kroah-Hartman <[email protected]>; Rafael J .
>>>> Wysocki <[email protected]>; [email protected]
>>>> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
>>>> Russell King <[email protected]>; Linus Walleij
>>>> <[email protected]>; Vinod Koul <[email protected]>; BOUGH CHEN
>>>> <[email protected]>; Ludovic Barre <[email protected]>;
>>>> [email protected]; [email protected]; Ulf
>>>> Hansson <[email protected]>
>>>> Subject: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus
>>>> level
>>>>
>>>> It's currently the amba/platform driver's responsibility to initialize
>>>> the pointer, dma_parms, for its corresponding struct device. The
>>>> benefit with this approach allows us to avoid the initialization and
>>>> to not waste memory for the struct device_dma_parameters, as this can
>>>> be decided on a case by case basis.
>>>>
>>>> However, it has turned out that this approach is not very practical.
>>>> Not only does it lead to open coding, but also to real errors. In
>>>> principle callers of
>>>> dma_set_max_seg_size() doesn't check the error code, but just assumes
>>>> it succeeds.
>>>>
>>>> For these reasons, this series initializes the dma_parms from the
>>>> amba/platform bus at the device registration point. This also follows
>>>> the way the PCI devices are being managed, see pci_device_add().
>>>>
>>>> If it turns out that this is an acceptable solution, we probably also
>>>> want the changes for stable, but I am not sure if it applies without conflicts.
>>>>
>>>> The series is based on v5.6-rc7.
>>>>
>>>
>>> Hi Ulf,
>>>
>>> Since i.MXQM SMMU related code still not upstream yet, so I apply your
>>> patches on our internal Linux branch based on v5.4.24, and find it do not work
>>> on my side. Maybe for platform core drivers, there is a gap between v5.4.24
>>> and v5.6-rc7 which has the impact.
>>> I will try to put our SMMU related code into v5.6-rc7, then do the test again.
>>>
>>>
>>
>> Hi Ulf,
>>
>> On the latest Linux-next branch, the top commit 89295c59c1f063b533d071ca49d0fa0c0783ca6f (tag: next-20200326), after add your two patches, I just add the simple debug code as following in the /driver/mmc/core/queue.c, but seems still not work as our expect, logically, it should work, so can you or anyone test on other platform? This seems weird.
>
> You are right, this doesn't work for platform devices being added
> through the OF path.
>
> In other words, of_platform_device_create_pdata() manually allocates
> the platform device and assigns it the &platform_bus_type, but without
> calling platform_device_add().
>
> For amba, it works fine, as in that OF path, amba_device_add() is called. Hmm.
>
> I re-spin this, to address the problem. Perhaps we simply need to use
> the ->probe() path.
FWIW we already have setup_pdev_dma_masks(), so it might be logical to
include dma_parms in there too.
Robin.
>
> Kind regards
> Uffe
>
>>
>> diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
>> index 25bee3daf9e2..f091280f7ffb 100644
>> --- a/drivers/mmc/core/queue.c
>> +++ b/drivers/mmc/core/queue.c
>> @@ -403,6 +403,13 @@ static void mmc_setup_queue(struct mmc_queue *mq, struct mmc_card *card)
>> blk_queue_max_segment_size(mq->queue,
>> round_down(host->max_seg_size, block_size));
>>
>> + pr_err("###### the max segment size is %d\n", queue_max_segment_size(mq->queue));
>> + if (host->parent->dma_parms) {
>> + pr_err("######### the dma parms has value\n");
>> + } else if (!(host->parent->dma_parms)) {
>> + pr_err("######## the dma parms is zero!!\n");
>> + }
>> +
>> dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue));
>>
>> INIT_WORK(&mq->recovery_work, mmc_mq_recovery_handler);
>>
>> Here is the log I got when system run, even after your patch, the dev->dma_parms is still NULL.
>> [ 0.989853] mmc0: new HS400 MMC card at address 0001
>> [ 0.995708] sdhci-esdhc-imx 30b50000.mmc: Got CD GPIO
>> [ 0.999374] ###### the max segment size is 65024
>> [ 1.008594] ######## the dma parms is zero!!
>> [ 1.012875] mmcblk0: mmc0:0001 IB2932 29.2 GiB
>> [ 1.017569] ###### the max segment size is 65024
>> [ 1.022195] ######## the dma parms is zero!!
>> [ 1.026479] mmcblk0boot0: mmc0:0001 IB2932 partition 1 4.00 MiB
>> [ 1.032541] ###### the max segment size is 65024
>> [ 1.035198] mmc1: SDHCI controller on 30b50000.mmc [30b50000.mmc] using ADMA
>> [ 1.037169] ######## the dma parms is zero!!
>> [ 1.048493] mmcblk0boot1: mmc0:0001 IB2932 partition 2 4.00 MiB
>> [ 1.054531] mmcblk0rpmb: mmc0:0001 IB2932 partition 3 4.00 MiB, chardev (234:0)
>>
>>
>> Regards
>> Haibo Chen
>>> Best Regards
>>> Haibo Chen
>>>
>>>> Kind regards
>>>> Ulf Hansson
>>>>
>>>> Ulf Hansson (2):
>>>> driver core: platform: Initialize dma_parms for platform devices
>>>> amba: Initialize dma_parms for amba devices
>>>>
>>>> drivers/amba/bus.c | 2 ++
>>>> drivers/base/platform.c | 1 +
>>>> include/linux/amba/bus.h | 1 +
>>>> include/linux/platform_device.h | 1 +
>>>> 4 files changed, 5 insertions(+)
>>>>
>>>> --
>>>> 2.20.1
>>
>
> _______________________________________________
> linux-arm-kernel mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
On Fri, 27 Mar 2020 at 20:15, Robin Murphy <[email protected]> wrote:
>
> On 2020-03-27 3:34 pm, Ulf Hansson wrote:
> > On Fri, 27 Mar 2020 at 04:02, BOUGH CHEN <[email protected]> wrote:
> >>
> >>
> >>> -----Original Message-----
> >>> From: BOUGH CHEN
> >>> Sent: 2020年3月26日 12:41
> >>> To: Ulf Hansson <[email protected]>; Greg Kroah-Hartman
> >>> <[email protected]>; Rafael J . Wysocki <[email protected]>;
> >>> [email protected]
> >>> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> >>> Russell King <[email protected]>; Linus Walleij <[email protected]>;
> >>> Vinod Koul <[email protected]>; Ludovic Barre <[email protected]>;
> >>> [email protected]; [email protected]
> >>> Subject: RE: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus level
> >>>
> >>>> -----Original Message-----
> >>>> From: Ulf Hansson <[email protected]>
> >>>> Sent: 2020年3月25日 19:34
> >>>> To: Greg Kroah-Hartman <[email protected]>; Rafael J .
> >>>> Wysocki <[email protected]>; [email protected]
> >>>> Cc: Arnd Bergmann <[email protected]>; Christoph Hellwig <[email protected]>;
> >>>> Russell King <[email protected]>; Linus Walleij
> >>>> <[email protected]>; Vinod Koul <[email protected]>; BOUGH CHEN
> >>>> <[email protected]>; Ludovic Barre <[email protected]>;
> >>>> [email protected]; [email protected]; Ulf
> >>>> Hansson <[email protected]>
> >>>> Subject: [PATCH 0/2] amba/platform: Initialize dma_parms at the bus
> >>>> level
> >>>>
> >>>> It's currently the amba/platform driver's responsibility to initialize
> >>>> the pointer, dma_parms, for its corresponding struct device. The
> >>>> benefit with this approach allows us to avoid the initialization and
> >>>> to not waste memory for the struct device_dma_parameters, as this can
> >>>> be decided on a case by case basis.
> >>>>
> >>>> However, it has turned out that this approach is not very practical.
> >>>> Not only does it lead to open coding, but also to real errors. In
> >>>> principle callers of
> >>>> dma_set_max_seg_size() doesn't check the error code, but just assumes
> >>>> it succeeds.
> >>>>
> >>>> For these reasons, this series initializes the dma_parms from the
> >>>> amba/platform bus at the device registration point. This also follows
> >>>> the way the PCI devices are being managed, see pci_device_add().
> >>>>
> >>>> If it turns out that this is an acceptable solution, we probably also
> >>>> want the changes for stable, but I am not sure if it applies without conflicts.
> >>>>
> >>>> The series is based on v5.6-rc7.
> >>>>
> >>>
> >>> Hi Ulf,
> >>>
> >>> Since i.MXQM SMMU related code still not upstream yet, so I apply your
> >>> patches on our internal Linux branch based on v5.4.24, and find it do not work
> >>> on my side. Maybe for platform core drivers, there is a gap between v5.4.24
> >>> and v5.6-rc7 which has the impact.
> >>> I will try to put our SMMU related code into v5.6-rc7, then do the test again.
> >>>
> >>>
> >>
> >> Hi Ulf,
> >>
> >> On the latest Linux-next branch, the top commit 89295c59c1f063b533d071ca49d0fa0c0783ca6f (tag: next-20200326), after add your two patches, I just add the simple debug code as following in the /driver/mmc/core/queue.c, but seems still not work as our expect, logically, it should work, so can you or anyone test on other platform? This seems weird.
> >
> > You are right, this doesn't work for platform devices being added
> > through the OF path.
> >
> > In other words, of_platform_device_create_pdata() manually allocates
> > the platform device and assigns it the &platform_bus_type, but without
> > calling platform_device_add().
> >
> > For amba, it works fine, as in that OF path, amba_device_add() is called. Hmm.
> >
> > I re-spin this, to address the problem. Perhaps we simply need to use
> > the ->probe() path.
>
> FWIW we already have setup_pdev_dma_masks(), so it might be logical to
> include dma_parms in there too.
Yep, thanks for the suggestion. This work fine.
[...]
Kind regards
Uffe