linux-kernel - Re: [PATCH 0/3] mtd: core: protect access to mtd devices while in suspend

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

Message-ID: <20211015082206.244a2316@xps13>
Date:   Fri, 15 Oct 2021 08:22:06 +0200
From:   Miquel Raynal <miquel.raynal@...tlin.com>
To:     Boris Brezillon <boris.brezillon@...labora.com>
Cc:     Sean Nyekjaer <sean@...nix.com>,
        Richard Weinberger <richard@....at>,
        Vignesh Raghavendra <vigneshr@...com>,
        Boris Brezillon <bbrezillon@...nel.org>,
        linux-mtd@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 0/3] mtd: core: protect access to mtd devices while in
 suspend

Hi Sean,

boris.brezillon@...labora.com wrote on Mon, 11 Oct 2021 16:05:46 +0200:

> On Mon, 11 Oct 2021 13:52:50 +0200
> Sean Nyekjaer <sean@...nix.com> wrote:
> 
> > Follow-up on discussion in https://lkml.org/lkml/2021/10/4/41
> > 
> > Changes since from rfc v1/v2:
> >  - added access protection for all device access hooks in mtd_info.
> >  - added Suggested-by to [1/3] patch.
> >  - removed refereces to commit ef347c0cfd61 ("mtd: rawnand: gpmi: Implement exec_op")
> >    from commit msg as commit 013e6292aaf5 ("mtd: rawnand: Simplify the locking") is 
> >    to be blamed.
> >  - tested on a kernel with LOCKDEP enabled.
> > 
> > @Miquel: I havn't covered every ioctl, to me it looks like they havn't
> > direct device access.

Yes indeed it looks like they re-use most of the mtdcore.c functions,
so it should be fine.

> > One (small) issue still present. gpmi_nand.c uses the rwsem before it's
> > initialized. Seems cumbersome to have every mtd/nand driver to call
> > init_waitqueue_head() and init_rwsem(). Could we somehow move the call
> > to mtd_set_dev_defaults() before nand_create_bbt()?  
> 
> I have a nasty trick for that one, but I'm not sure Miquel will like it
> (actually, I don't like it either, but it's so simple compared to the
> other options we have that I'm tempted to go for this approach until
> someone has time to invest in a cleaner solution :-)):
> 
> diff --git a/drivers/mtd/nand/raw/nand_base.c b/drivers/mtd/nand/raw/nand_base.c
> index 3d6c6e880520..a9ac2d528a4d 100644
> --- a/drivers/mtd/nand/raw/nand_base.c
> +++ b/drivers/mtd/nand/raw/nand_base.c
> @@ -6222,8 +6222,6 @@ static int nand_scan_tail(struct nand_chip *chip)
>         mtd->_sync = nand_sync;
>         mtd->_lock = nand_lock;
>         mtd->_unlock = nand_unlock;
> -       mtd->_suspend = nand_suspend;
> -       mtd->_resume = nand_resume;
>         mtd->_reboot = nand_shutdown;
>         mtd->_block_isreserved = nand_block_isreserved;
>         mtd->_block_isbad = nand_block_isbad;
> @@ -6269,6 +6267,13 @@ static int nand_scan_tail(struct nand_chip *chip)
>         if (ret)
>                 goto err_free_secure_regions;
>  
> +       /*
> +        * Populate the suspend/resume hooks after the BBT has been scanned to
> +        * avoid using the suspend lock and resume waitqueue which are only
> +        * initialized when mtd_device_register() is called.
> +        */
> +       mtd->_suspend = nand_suspend;
> +       mtd->_resume = nand_resume;
>         return 0;

I'm fine with this as long as it is documented for now.

Thanks,
Miquèl