Re: [PATCH 2/3] mm: cma: allocation trigger

[Sasha Levin <sasha.levin@oracle.com>]

On 01/22/2015 03:26 AM, Joonsoo Kim wrote:
> On Tue, Jan 20, 2015 at 12:38:32PM -0500, Sasha Levin wrote:
>> Provides a userspace interface to trigger a CMA allocation.
>>
>> Usage:
>>
>> 	echo [pages] > alloc
>>
>> This would provide testing/fuzzing access to the CMA allocation paths.
>>
>> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
>> ---
>>  mm/cma_debug.c |   58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 58 insertions(+)
>>
>> diff --git a/mm/cma_debug.c b/mm/cma_debug.c
>> index 3a25413..eda0a41 100644
>> --- a/mm/cma_debug.c
>> +++ b/mm/cma_debug.c
>> @@ -7,9 +7,22 @@
>>  
>>  #include <linux/debugfs.h>
>>  #include <linux/cma.h>
>> +#include <linux/list.h>
>> +#include <linux/kernel.h>
>> +#include <linux/dma-contiguous.h>
>> +#include <linux/slab.h>
>>  
>>  #include "cma.h"
>>  
>> +struct cma_mem {
>> +	struct hlist_node node;
>> +	struct page *p;
>> +	unsigned long n;
>> +};
>> +
>> +static HLIST_HEAD(cma_mem_head);
>> +static DEFINE_SPINLOCK(cma_mem_head_lock);
>> +
>>  static struct dentry *cma_debugfs_root;
>>  
>>  static int cma_debugfs_get(void *data, u64 *val)
>> @@ -44,6 +57,48 @@ static void cma_debugfs_add_one(struct cma *cma, int idx)
>>  	debugfs_create_u32_array("bitmap", S_IRUGO, tmp, (u32*)cma->bitmap, u32s);
>>  }
>>  
>> +static void cma_add_to_cma_mem_list(struct cma_mem *mem)
>> +{
>> +	spin_lock(&cma_mem_head_lock);
>> +	hlist_add_head(&mem->node, &cma_mem_head);
>> +	spin_unlock(&cma_mem_head_lock);
>> +}
>> +
>> +static int cma_alloc_mem(int count)
>> +{
>> +	struct cma_mem *mem;
>> +	struct page *p;
>> +
>> +	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
>> +	if (!mem) 
>> +		return -ENOMEM;
>> +
>> +	p = cma_alloc(dma_contiguous_default_area, count, CONFIG_CMA_ALIGNMENT);
>> +	if (!p) {
>> +		pr_info("CMA: Allocation failed.\n");
>> +		kfree(mem);
>> +		return -ENOMEM;
>> +	}
> 
> Hello,
> 
> I'm glad to see this patchset. I had similar one privately. :)
> Without this kind of facility, testing CMA is really hard.
> 
> Anyway, dma_contiguous_default_area is defined only in CONFIG_DMA_CMA.
> So, it would cause build break.
> 
> And, I think that it is better for this tester to have an ability to
> allocate pages on specific CMA reserved region. It could be easily
> achieved by moving 'alloc' registration to cma_debugfs_add_one(). Is
> there any reason not to do that?

Agreed, that would make more sense.

>> +
>> +	pr_info("CMA: Allocated %d pages at %p\n", count, p);
> 
> I don't think this output is needed. If CONFIG_CMA_DEBUG is enabled,
> cma_alloc() print similar output.

I didn't see any other output (and I have CONFIG_CMA_DEBUG set), which
is why I've added this pr_info().


Thanks,
Sasha