Re: [PATCH] mm, dax: VMA with vm_ops->pfn_mkwrite wants to be write-notified

[Boaz Harrosh <boaz@plexistor.com>]

On 09/01/2015 02:10 PM, Kirill A. Shutemov wrote:
> On Tue, Sep 01, 2015 at 01:54:26PM +0300, Boaz Harrosh wrote:
>> On 09/01/2015 01:22 PM, Kirill A. Shutemov wrote:
>>> For VM_PFNMAP and VM_MIXEDMAP we use vm_ops->pfn_mkwrite instead of
>>> vm_ops->page_mkwrite to notify abort write access. This means we want
>>> vma->vm_page_prot to be write-protected if the VMA provides this vm_ops.
>>>
>>
>> Hi Kirill
>>
>> I will test with this right away and ACK on this.
>>
>> Hmm so are you saying we might be missing some buffer modifications right now.
>>
>> What would be a theoretical scenario that will cause these missed events?
> 
> On writable mapping with vm_ops->pfn_mkwrite, but without
> vm_ops->page_mkwrite: read fault followed by write access to the pfn.
> Writable pte will be set up on read fault and write fault will not be
> generated.
> 
> I found it examining Dave's complain on generic/080:
> 
> http://lkml.kernel.org/g/20150831233803.GO3902@dastard
> 
> Although I don't think it's the reason.
> 
>> I would like to put a test in our test rigs that should fail today and this
>> patch fixes.
>>
>> [In our system every modified pmem block is also RDMAed to a remote
>>  pmem for HA, a missed modification will make the two copies unsynced]
> 
> It shouldn't be a problem for ext2/ext4 as they provide both pfn_mkwrite
> and page_mkwrite.
> 

Ha right we have both as well, and so should xfs I think (because of the
zero pages thing, in fact any dax.c user should).

Thanks so this verifies why we could not see any such breakage.

ACK-by: Boaz Harrosh <boaz@plexistor.com>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>