linux-kernel - Re: [PATCH v3] kernel/resource: Fix locking in request_free_mem

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <4738a7e1-ec62-0b11-28df-e81c0f904f84@nvidia.com>
Date:   Mon, 29 Mar 2021 22:17:41 -0700
From:   John Hubbard <jhubbard@...dia.com>
To:     Alistair Popple <apopple@...dia.com>
CC:     <akpm@...ux-foundation.org>, <linux-mm@...ck.org>,
        <linux-kernel@...r.kernel.org>, <david@...hat.com>,
        <daniel.vetter@...ll.ch>, <dan.j.williams@...el.com>,
        <gregkh@...uxfoundation.org>, <jglisse@...hat.com>,
        <bsingharora@...il.com>, kernel test robot <oliver.sang@...el.com>
Subject: Re: [PATCH v3] kernel/resource: Fix locking in
 request_free_mem_region

On 3/29/21 9:59 PM, Alistair Popple wrote:
...
>>>    		res->desc = IORES_DESC_DEVICE_PRIVATE_MEMORY;
>>> +		if (dev) {
>>> +			dr->parent = &iomem_resource;
>>> +			dr->start = addr;
>>> +			dr->n = size;
>>> +			devres_add(dev, dr);
>>> +		}
>>> +
>>> +		write_unlock(&resource_lock);
>>> +		revoke_iomem(res);
>>
>> This is new, and not mentioned in the commit log, and therefore quite
>> surprising. It seems like the right thing to do but it also seems like a
>> different fix! I'm not saying that it should be a separate patch, but it
>> does seem worth loudly mentioning in the commit log, yes?
> 
> This isn't a different fix though, it is just about maintaining the original
> behaviour which called revoke_iomem() after dropping the lock. I inadvertently
> switched this around in the initial patch such that revoke_iomem() got called
> with the lock, leading to deadlock on x86 with CONFIG_IO_STRICT_DEVMEM=y.
> 
> This does change the order of revoke_iomem() and devres_add() slightly, but as
> far as I can tell that shouldn't be an issue. Can call that out in the commit
> log.

Maybe that's why it looked like a change to me. I do think it's worth mentioning.


thanks,
-- 
John Hubbard
NVIDIA