linux-kernel - Re: [PATCH v3 3/5] nvme: add the NVME_ENABLE_MULTIPATH

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: <CADUfDZoNdUBuX06FB+jsduS=shgPTKxbr_dYdtS+adEe2hnVWg@mail.gmail.com>
Date: Mon, 31 Mar 2025 16:55:50 -0700
From: Caleb Sander Mateos <csander@...estorage.com>
To: John Meneghini <jmeneghi@...hat.com>
Cc: kbusch@...nel.org, hch@....de, sagi@...mberg.me, loberman@...hat.com, 
	linux-nvme@...ts.infradead.org, linux-kernel@...r.kernel.org, 
	emilne@...hat.com, bgurney@...hat.com
Subject: Re: [PATCH v3 3/5] nvme: add the NVME_ENABLE_MULTIPATH_PARAM config option

On Mon, Mar 31, 2025 at 7:29 AM John Meneghini <jmeneghi@...hat.com> wrote:
>
> On 3/28/25 3:49 PM, Caleb Sander Mateos wrote:
> >>   bool multipath = true;
> >> +#ifdef CONFIG_NVME_ENABLE_MULTIPATH_PARAM
> >>   module_param(multipath, bool, 0444);
> >>   MODULE_PARM_DESC(multipath,
> >>          "turn on native support for multiple controllers per subsystem");
> >> +#endif
> > If CONFIG_NVME_ENABLE_MULTIPATH_PARAM is disabled, could you #define
> > multipath false in place of the global variable? That would allow the
> > compiler to optimize out the multipath checks and the resulting dead
> > code.
>
> If we #define multipath to be false that would permanently disable the multipath code
> when CONFIG_NVME_ENABLE_MULTIPATH_PARAM=n, and this is not what we want. The purpose
> of CONFIG_NVME_ENABLE_MULTIPATH_PARAM=n is to simply remove the core_nvme.multipath
> parameter from the kernel so the user no longer has the ability to change
> the bool multipath variable.  We want multipath to be true.

Sorry, I just meant to #define multipath to a constant. I mixed up the
boolean value. If CONFIG_NVME_MULTIPATH is set, of course it should be
true, not false. If CONFIG_NVME_MULTIPATH is not set, multipath is
already defined as false.

My point was just that the compiler will pessimistically assume a
global variable with extern visibility could be written from some
other translation unit at any time. But
CONFIG_NVME_ENABLE_MULTIPATH_PARAM removes the only way for the global
variable to change at runtime, so it can just be a constant.

Best,
Caleb