Re: [PATCH v2 2/3] mm: cma: allocation trigger

[Joonsoo Kim <js1304@gmail.com>]

2015-01-28 0:08 GMT+09:00 Sasha Levin <sasha.levin@oracle.com>:
> On 01/27/2015 03:06 AM, Joonsoo Kim wrote:
>> On Mon, Jan 26, 2015 at 09:26:04AM -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 |   60 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
>>>  1 file changed, 58 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/mm/cma_debug.c b/mm/cma_debug.c
>>> index 3a25413..39c7116 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>
>>
>> dma-contiguous.h doesn't needed now.
>
> Right.
>
>>> +#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;
>>
>> How about keeping cma_mem_head on each cma area separately?
>
> Good point, we're mixing allocations here.
>
>>>  static int cma_debugfs_get(void *data, u64 *val)
>>> @@ -23,8 +36,48 @@ static int cma_debugfs_get(void *data, u64 *val)
>>>
>>>  DEFINE_SIMPLE_ATTRIBUTE(cma_debugfs_fops, cma_debugfs_get, NULL, "%llu\n");
>>>
>>> -static void cma_debugfs_add_one(struct cma *cma, int idx)
>>> +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(struct cma *cma, int count)
>>>  {
>>> +    struct cma_mem *mem;
>>> +    struct page *p;
>>> +
>>> +    mem = kzalloc(sizeof(*mem), GFP_KERNEL);
>>> +    if (!mem)
>>> +            return -ENOMEM;
>>> +
>>> +    p = cma_alloc(cma, count, CONFIG_CMA_ALIGNMENT);
>>> +    if (!p) {
>>> +            kfree(mem);
>>> +            return -ENOMEM;
>>> +    }
>>
>> CONFIG_CMA_ALIGNMENT looks not good. It means just maximum aligment
>> so it is odd to use this value in testing. Is there special meaning
>> to use it here?
>
> No good reason, I stole that from a different piece of code.
>
>> Could we also get alignment parameter from user? Something like below.
>>
>> echo "4 1" > alloc
>> 4 for number of pages
>> 1 for alignment.
>>
>> If it is impossible, just 0 looks better than CONFIG_CMA_ALIGNMENT.
>
> I'd rather keep it simple and use a single parameter for now.

Okay, then use 0 rather than CONFIG_CMA_ALIGNMENT.

Thanks.