linux-kernel - Re: [PATCH 2/4] gpiolib: cdev: allocate linereq using kvzalloc()

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]

Date: Wed, 20 Dec 2023 16:58:34 +0200
From: Andy Shevchenko <andy@...nel.org>
To: Kent Gibson <warthog618@...il.com>
Cc: linux-kernel@...r.kernel.org, linux-gpio@...r.kernel.org, brgl@...ev.pl,
	linus.walleij@...aro.org
Subject: Re: [PATCH 2/4] gpiolib: cdev: allocate linereq using kvzalloc()

On Wed, Dec 20, 2023 at 10:53:07PM +0800, Kent Gibson wrote:
> On Wed, Dec 20, 2023 at 04:30:24PM +0200, Andy Shevchenko wrote:
> > On Wed, Dec 20, 2023 at 09:51:04AM +0800, Kent Gibson wrote:
> > > The size of struct linereq may exceed a page, so allocate space for
> > > it using kvzalloc() instead of kzalloc().
> >
> > It might be this needs a bit of elaboration. The kmalloc() tries to allocate
> > a contiguous (in physical address space) chunk of memory and with fragmented
> > memory it might be not possible. So the above issue might (rarely) happen.
> > In most cases the call to kmalloc() will succeed.
> 
> For sure, the kzalloc() generally works - or we wouldn't've gotten this
> far as tests with MAX_LINES would've been failing.
> We are targetting a very niche failure mode here.
> 
> The size allocated can only be determined at runtime, may be more or
> less than a page, and we don't care whether the physical memory allocated
> is contiguous.
> As such kvzalloc() is the appropriate allocator.
> 
> Are you suggesting repeating the relevant sections of the
> kmalloc/vmalloc() documentation or Memory Allocation Guide as part of the
> checkin comment?

I suggesting to make clear in the commit message that:
- there is no bug per se with the code logic (i.o.w.
  there is no issue to have allocations bigger than one page)
- this is very rarely case when it might be a problem

You can also put a reference to the documentation, if you wish.
This should be harmless and adding not more than a line into
the commit message (or even as a Link: tag to the HTML variant of it).

-- 
With Best Regards,
Andy Shevchenko