[PATCH] lib/dma-debug: check for vmalloc buffers used with the DMA-API

[PATCH] lib/dma-debug: check for vmalloc buffers used with the DMA-API

[Jan Luebbe]

Signed-off-by: Jan Luebbe <jlu@pengutronix.de>
---
I'm not sure why we don't check for high memory, is there any problem
with this approach?

 lib/dma-debug.c |    3 +++
 1 file changed, 3 insertions(+)

diff --git a/lib/dma-debug.c b/lib/dma-debug.c
index 66ce414..7d60ff0 100644
--- a/lib/dma-debug.c
+++ b/lib/dma-debug.c
@@ -944,6 +944,9 @@ static void check_for_illegal_area(struct device *dev, void *addr, unsigned long
 	if (overlap(addr, len, _text, _etext) ||
 	    overlap(addr, len, __start_rodata, __end_rodata))
 		err_printk(dev, NULL, "DMA-API: device driver maps memory from kernel text or rodata [addr=%p] [len=%lu]\n", addr, len);
+	if (is_vmalloc_or_module_addr(addr) ||
+	    is_vmalloc_or_module_addr(addr+len))
+		err_printk(dev, NULL, "DMA-API: device driver maps memory from vmalloc or module address range [addr=%p] [len=%lu]\n", addr, len);
 }
 
 static void check_sync(struct device *dev,
-- 
1.7.10.4

Re: [PATCH] lib/dma-debug: check for vmalloc buffers used with the DMA-API

[Joerg Roedel]

On Mon, Sep 03, 2012 at 01:00:04PM +0200, Jan Luebbe wrote:
> Signed-off-by: Jan Luebbe <jlu@pengutronix.de>
> ---
> I'm not sure why we don't check for high memory, is there any problem
> with this approach?
> 
>  lib/dma-debug.c |    3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/lib/dma-debug.c b/lib/dma-debug.c
> index 66ce414..7d60ff0 100644
> --- a/lib/dma-debug.c
> +++ b/lib/dma-debug.c
> @@ -944,6 +944,9 @@ static void check_for_illegal_area(struct device *dev, void *addr, unsigned long
>  	if (overlap(addr, len, _text, _etext) ||
>  	    overlap(addr, len, __start_rodata, __end_rodata))
>  		err_printk(dev, NULL, "DMA-API: device driver maps memory from kernel text or rodata [addr=%p] [len=%lu]\n", addr, len);
> +	if (is_vmalloc_or_module_addr(addr) ||
> +	    is_vmalloc_or_module_addr(addr+len))
> +		err_printk(dev, NULL, "DMA-API: device driver maps memory from vmalloc or module address range [addr=%p] [len=%lu]\n", addr, len);

Hmm, this only checks whether the start or end+1 address is in the
vmalloc/module range. I think it is better to check for overlap against
VMALLOC_START/END and do a seperate overlap check for MODULES_VADDR and
MODULES_END when needed.

Thanks,

	Joerg

-- 
AMD Operating System Research Center

Advanced Micro Devices GmbH Einsteinring 24 85609 Dornach
General Managers: Alberto Bozzo
Registration: Dornach, Landkr. Muenchen; Registerger. Muenchen, HRB Nr. 43632

[PATCH v2] lib/dma-debug: check for vmalloc and module addresses used with the DMA-API

[Jan Luebbe]

Signed-off-by: Jan Luebbe <jlu@pengutronix.de>
---

Changes since v1:
 - As recommended by Joerg Roedel, check for overlap with vmalloc and
   module address ranges.

 lib/dma-debug.c |    8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/lib/dma-debug.c b/lib/dma-debug.c
index 66ce414..b35992b 100644
--- a/lib/dma-debug.c
+++ b/lib/dma-debug.c
@@ -944,6 +944,14 @@ static void check_for_illegal_area(struct device *dev, void *addr, unsigned long
 	if (overlap(addr, len, _text, _etext) ||
 	    overlap(addr, len, __start_rodata, __end_rodata))
 		err_printk(dev, NULL, "DMA-API: device driver maps memory from kernel text or rodata [addr=%p] [len=%lu]\n", addr, len);
+#ifdef CONFIG_MMU
+	if (overlap(addr, len, (void *)VMALLOC_START, (void *)VMALLOC_END))
+		err_printk(dev, NULL, "DMA-API: device driver maps memory from vmalloc address range [addr=%p] [len=%lu]\n", addr, len);
+#endif
+#if defined(CONFIG_MODULES) && defined(MODULES_VADDR)
+	if (overlap(addr, len, (void *)MODULES_VADDR, (void *)MODULES_END))
+		err_printk(dev, NULL, "DMA-API: device driver maps memory from kernel module address range [addr=%p] [len=%lu]\n", addr, len);
+#endif
 }
 
 static void check_sync(struct device *dev,
-- 
1.7.10.4