[RFC PATCH v6.6] maple_tree: Fix MA_STATE_PREALLOC flag in mas_preallocate()

[Hailong Liu]

On 5/9/2025 11:27 PM, Suren Baghdasaryan wrote:
> On Wed, May 7, 2025 at 8:50 AM Liam R. Howlett <Liam.Howlett at oracle.com> wrote:
>>
>> * Liam R. Howlett <Liam.Howlett at oracle.com> [250428 21:48]:
>>> Temporarily clear the preallocation flag when explicitly requesting
>>> allocations.  Pre-existing allocations are already counted against the
>>> request through mas_node_count_gfp(), but the allocations will not
>>> happen if the MA_STATE_PREALLOC flag is set.  This flag is meant to
>>> avoid re-allocating in bulk allocation mode, and to detect issues with
>>> preallocation calculations.
>>>
>>> The MA_STATE_PREALLOC flag should also always be set on zero allocations
>>> so that detection of underflow allocations will print a WARN_ON() during
>>> consumption.
>>>
>>> User visible effect of this flaw is a WARN_ON() followed by a null
>>> pointer dereference when subsequent requests for larger number of nodes
>>> is ignored, such as the vma merge retry in mmap_region() caused by
>>> drivers altering the vma flags.
>>>
>>> Reported-by: Zhaoyang Huang <zhaoyang.huang at unisoc.com>
>>> Reported-by: Hailong Liu <hailong.liu at oppo.com>
>>> Fixes: 54a611b605901 ("Maple Tree: add new data structure")
>>> Link: https://lore.kernel.org/all/1652f7eb-a51b-4fee-8058-c73af63bacd1@oppo.com/
>>> Link: https://lore.kernel.org/all/20250428184058.1416274-1-Liam.Howlett@oracle.com/
>>> Cc: Lorenzo Stoakes <lorenzo.stoakes at oracle.com>
>>> Cc: Suren Baghdasaryan <surenb at google.com>
>>> Cc: Hailong Liu <hailong.liu at oppo.com>
>>> Cc: zhangpeng.00 at bytedance.com <zhangpeng.00 at bytedance.com>
>>> Cc: Steve Kang <Steve.Kang at unisoc.com>
>>> Cc: Matthew Wilcox <willy at infradead.org>
>>> Cc: Sidhartha Kumar <sidhartha.kumar at oracle.com>
>>> Signed-off-by: Liam R. Howlett <Liam.Howlett at oracle.com>
>>
>> ...
>>
>> I have a version of this for mm-new and I'd like to send it out.  Once
>> this is upstream, it will be backported to the stable kernels with
>> something that looks a lot like what I sent out here.
>>
>> Did this fix the issue in the longer running tests?
> 
> - everyone else
> 
> Hi Liam,
> I think the delay is due to the holidays in China. I requested an
> update from the partners but they will probably provide it next week.
Sorry for late reply. We applied this patch and verified it fix the issue.

Feel free to add

Tested-by: Hailong Liu <hailong.liu at oppo.com>

Thanks,
Hailong.

> Thanks,
> Suren.
> 
>>
>> Thanks,
>> Liam