This patch fixes a build failure introduced by the patch
atmel-mci: change use of dma slave interface by Nicolas Ferre
by changing mci_dma_slave to the correct name of mci_dma_data
This should make the avr32 tree build again.
References:
http://kisskb.ellerman.id.au/kisskb/buildresult/1893610/
http://git.kernel.org/?p=linux/kernel/git/sfr/linux-next.git;a=commitdiff;h=2635d1ba711560d521f6218c585a3e0401f566e1
Patch against Linus' tree.
Signed-off-by: Peter Huewe <[email protected]>
---
arch/avr32/mach-at32ap/at32ap700x.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/avr32/mach-at32ap/at32ap700x.c b/arch/avr32/mach-at32ap/at32ap700x.c
index 1aa1ea5..7d4ed42 100644
--- a/arch/avr32/mach-at32ap/at32ap700x.c
+++ b/arch/avr32/mach-at32ap/at32ap700x.c
@@ -1325,7 +1325,7 @@ struct platform_device *__init
at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
{
struct platform_device *pdev;
- struct mci_dma_slave *slave;
+ struct mci_dma_data *slave;
u32 pioa_mask;
u32 piob_mask;
@@ -1344,7 +1344,7 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
ARRAY_SIZE(atmel_mci0_resource)))
goto fail;
- slave = kzalloc(sizeof(struct mci_dma_slave), GFP_KERNEL);
+ slave = kzalloc(sizeof(struct mci_dma_data), GFP_KERNEL);
slave->sdata.dma_dev = &dw_dmac0_device.dev;
slave->sdata.reg_width = DW_DMA_SLAVE_WIDTH_32BIT;
--
1.6.4.4
On Fri, 25 Dec 2009 20:08:45 +0100
Peter Huewe <[email protected]> wrote:
> This patch fixes a build failure introduced by the patch
> atmel-mci: change use of dma slave interface by Nicolas Ferre
> by changing mci_dma_slave to the correct name of mci_dma_data
>
> This should make the avr32 tree build again.
>
Ah, great, I have not gotten around to test this yet. Thanks for the
patch. I've done some minor changes and also fixed the kzalloc/kfree
error handling. Please see the inlined patch below.
>From c8d4b6c01b19044467867b24e218b8a59ed04291 Mon Sep 17 00:00:00 2001
From: Hans-Christian Egtvedt <[email protected]>
Date: Mon, 28 Dec 2009 09:08:00 +0100
Subject: [PATCH] avr32: Fix build failure when using atmel-mci caused by typo
This patch fixes a build failure introduced by the patch atmel-mci: change use
of dma slave interface by Nicolas Ferre by changing mci_dma_slave to the
correct name of mci_dma_data.
It also will only free the struct mci_dma_data if it has been allocated.
This should make the avr32 tree build again.
References:
http://kisskb.ellerman.id.au/kisskb/buildresult/1893610/
http://git.kernel.org/?p=linux/kernel/git/sfr/linux-next.git;a=commitdiff;h=2635d1ba711560d521f6218c585a3e0401f566e1
Patch against Linus' tree.
Signed-off-by: Peter Huewe <[email protected]>
Signed-off-by: Hans-Christian Egtvedt <[email protected]>
---
arch/avr32/mach-at32ap/at32ap700x.c | 13 +++++++------
1 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/arch/avr32/mach-at32ap/at32ap700x.c b/arch/avr32/mach-at32ap/at32ap700x.c
index 1aa1ea5..6e093c3 100644
--- a/arch/avr32/mach-at32ap/at32ap700x.c
+++ b/arch/avr32/mach-at32ap/at32ap700x.c
@@ -1325,7 +1325,7 @@ struct platform_device *__init
at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
{
struct platform_device *pdev;
- struct mci_dma_slave *slave;
+ struct mci_dma_data *slave;
u32 pioa_mask;
u32 piob_mask;
@@ -1344,7 +1344,7 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
ARRAY_SIZE(atmel_mci0_resource)))
goto fail;
- slave = kzalloc(sizeof(struct mci_dma_slave), GFP_KERNEL);
+ slave = kzalloc(sizeof(struct mci_dma_data), GFP_KERNEL);
slave->sdata.dma_dev = &dw_dmac0_device.dev;
slave->sdata.reg_width = DW_DMA_SLAVE_WIDTH_32BIT;
@@ -1357,7 +1357,7 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
if (platform_device_add_data(pdev, data,
sizeof(struct mci_platform_data)))
- goto fail;
+ goto fail_free;
/* CLK line is common to both slots */
pioa_mask = 1 << 10;
@@ -1381,7 +1381,7 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
/* Slot is unused */
break;
default:
- goto fail;
+ goto fail_free;
}
select_peripheral(PIOA, pioa_mask, PERIPH_A, 0);
@@ -1408,7 +1408,7 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
break;
default:
if (!data->slot[0].bus_width)
- goto fail;
+ goto fail_free;
data->slot[1].bus_width = 0;
break;
@@ -1419,9 +1419,10 @@ at32_add_device_mci(unsigned int id, struct mci_platform_data *data)
platform_device_add(pdev);
return pdev;
+fail_free:
+ kfree(slave);
fail:
data->dma_slave = NULL;
- kfree(slave);
platform_device_put(pdev);
return NULL;
}
--
1.6.0.4
--
Best regards,
Hans-Christian Egtvedt
Am Montag 28 Dezember 2009 09:13:18 schrieb Hans-Christian Egtvedt:
> On Fri, 25 Dec 2009 20:08:45 +0100
>
> Peter Huewe <[email protected]> wrote:
> > This patch fixes a build failure introduced by the patch
> > atmel-mci: change use of dma slave interface by Nicolas Ferre
> > by changing mci_dma_slave to the correct name of mci_dma_data
> >
> > This should make the avr32 tree build again.
>
> Ah, great, I have not gotten around to test this yet. Thanks for the
> patch. I've done some minor changes and also fixed the kzalloc/kfree
> error handling. Please see the inlined patch below.
>
> @@ -1344,7 +1344,7 @@ at32_add_device_mci(unsigned int id, struct
> mci_platform_data *data) ARRAY_SIZE(atmel_mci0_resource)))
> goto fail;
>
> - slave = kzalloc(sizeof(struct mci_dma_slave), GFP_KERNEL);
> + slave = kzalloc(sizeof(struct mci_dma_data), GFP_KERNEL);
>
> slave->sdata.dma_dev = &dw_dmac0_device.dev;
> slave->sdata.reg_width = DW_DMA_SLAVE_WIDTH_32BIT;
Great, thanks for the update.
The only thing that still catches my eye is that the kzalloc line can fail, so
you should perhaps add something like
if (!slave) {
printk(KERN_ERR "No memory left for at32ap700x: at32_add_device_mc");
goto fail;
}
(not tested :)
Apart from that:
Acked-By: Peter Huewe <[email protected]>
Thanks and best regards,
Peter
On Mon, 28 Dec 2009 10:59:30 +0100
Peter Hüwe <[email protected]> wrote:
> Am Montag 28 Dezember 2009 09:13:18 schrieb Hans-Christian Egtvedt:
> > On Fri, 25 Dec 2009 20:08:45 +0100
> >
> > Peter Huewe <[email protected]> wrote:
> > > This patch fixes a build failure introduced by the patch
> > > atmel-mci: change use of dma slave interface by Nicolas Ferre
> > > by changing mci_dma_slave to the correct name of mci_dma_data
> > >
> > > This should make the avr32 tree build again.
> >
> > Ah, great, I have not gotten around to test this yet. Thanks for the
> > patch. I've done some minor changes and also fixed the kzalloc/kfree
> > error handling. Please see the inlined patch below.
> >
>
> > @@ -1344,7 +1344,7 @@ at32_add_device_mci(unsigned int id, struct
> > mci_platform_data *data) ARRAY_SIZE(atmel_mci0_resource)))
> > goto fail;
> >
> > - slave = kzalloc(sizeof(struct mci_dma_slave), GFP_KERNEL);
> > + slave = kzalloc(sizeof(struct mci_dma_data), GFP_KERNEL);
> >
> > slave->sdata.dma_dev = &dw_dmac0_device.dev;
> > slave->sdata.reg_width = DW_DMA_SLAVE_WIDTH_32BIT;
>
>
> Great, thanks for the update.
> The only thing that still catches my eye is that the kzalloc line can fail, so
> you should perhaps add something like
> if (!slave) {
> printk(KERN_ERR "No memory left for at32ap700x: at32_add_device_mc");
> goto fail;
> }
>
Sounds like the patch should be split in two, one addressing the proper
naming of the struct mci_dma_data and one fixing the allocation of
memory and proper cleanup.
--
Best regards,
Hans-Christian Egtvedt
Peter Hüwe <[email protected]> wrote:
> Great, thanks for the update.
> The only thing that still catches my eye is that the kzalloc line can fail, so
> you should perhaps add something like
> if (!slave) {
> printk(KERN_ERR "No memory left for at32ap700x: at32_add_device_mc");
> goto fail;
> }
I agree, except that this is extremely unlikely to happen, so we should
either drop the printk() or use pr_debug() in order to waste as little
memory as possible in production systems.
I think your first patch is good as it is, so I will apply it. Please
send any additional fixes as a separate patch.
Haavard
Am Montag 28 Dezember 2009 11:20:28 schrieb Haavard Skinnemoen:
> Peter Hüwe <[email protected]> wrote:
> > Great, thanks for the update.
> > The only thing that still catches my eye is that the kzalloc line can
> > fail, so you should perhaps add something like
> > if (!slave) {
> > printk(KERN_ERR "No memory left for at32ap700x: at32_add_device_mc");
> > goto fail;
> > }
>
> I agree, except that this is extremely unlikely to happen, so we should
> either drop the printk() or use pr_debug() in order to waste as little
> memory as possible in production systems.
ok.
>
> I think your first patch is good as it is, so I will apply it. Please
> send any additional fixes as a separate patch.
Okay thanks for applying.
@Hans: can you create the cleanup patch or should I ?
Regards,
Peter
On Mon, 28 Dec 2009 11:34:36 +0100
Peter Hüwe <[email protected]> wrote:
> Am Montag 28 Dezember 2009 11:20:28 schrieb Haavard Skinnemoen:
> > Peter Hüwe <[email protected]> wrote:
> > > Great, thanks for the update.
> > > The only thing that still catches my eye is that the kzalloc line can
> > > fail, so you should perhaps add something like
> > > if (!slave) {
> > > printk(KERN_ERR "No memory left for at32ap700x: at32_add_device_mc");
> > > goto fail;
> > > }
> >
> > I agree, except that this is extremely unlikely to happen, so we should
> > either drop the printk() or use pr_debug() in order to waste as little
> > memory as possible in production systems.
> ok.
>
I skipped the print message, as no other platform devices will print
similar information. If people really need this kind of information
they could provide a patch adding it for all the platform devices.
> > I think your first patch is good as it is, so I will apply it. Please
> > send any additional fixes as a separate patch.
>
> Okay thanks for applying.
>
> @Hans: can you create the cleanup patch or should I ?
>
I'll do it, should arrive here shortly.
--
Best regards,
Hans-Christian Egtvedt
On Mon, 28 Dec 2009 12:07:23 +0100
Hans-Christian Egtvedt <[email protected]> wrote:
> On Mon, 28 Dec 2009 11:34:36 +0100
> Peter Hüwe <[email protected]> wrote:
<snipp>
> > > I think your first patch is good as it is, so I will apply it. Please
> > > send any additional fixes as a separate patch.
> >
> > Okay thanks for applying.
> >
> > @Hans: can you create the cleanup patch or should I ?
> >
>
> I'll do it, should arrive here shortly.
>
"here", by here I actually meant to the [email protected] mailing
list[1]. So I hope Håvard will snap it from there.
Sadly I just realized that the archives has been non functional since
mid November due to server misconfiguration. Fixed now.
1: http://lists.avr32linux.org/pipermail/kernel/2009-December/005484.html
--
Best regards,
Hans-Christian Egtvedt