linux-kernel - Re: [PATCH] mvsas:Fix possible NULL pointer deference in mvs_dev_found

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

Date:	Wed, 9 Mar 2016 22:33:47 +0200
From:	Dāvis Mosāns <davispuh@...il.com>
To:	Nicholas Krause <xerofoify@...il.com>
Cc:	James Bottomley <James.Bottomley@...senpartnership.com>,
	Johannes Thumshirn <jthumshirn@...e.de>,
	luisbg@....samsung.com, Wilfried.Weissmann@....at,
	"linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] mvsas:Fix possible NULL pointer deference in mvs_dev_found_notify

2016-03-09 15:58 GMT+02:00 Nicholas Krause <xerofoify@...il.com>:
> This adds properly checking after the call to mvs_find_dev_mvi
> due to this function being able to return a NULL pointer and if
> this does arise we will deference it in mvs_alloc_dev due to
> this function never checking if a NULL pointer is given as
> it's input argument.
>
> Signed-off-by: Nicholas Krause <xerofoify@...il.com>
> ---
>  drivers/scsi/mvsas/mv_sas.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/drivers/scsi/mvsas/mv_sas.c b/drivers/scsi/mvsas/mv_sas.c
> index 83cd3ea..7afb248 100644
> --- a/drivers/scsi/mvsas/mv_sas.c
> +++ b/drivers/scsi/mvsas/mv_sas.c
> @@ -1191,6 +1191,10 @@ int mvs_dev_found_notify(struct domain_device *dev, int lock)
>         struct mvs_device *mvi_device;
>
>         mvi = mvs_find_dev_mvi(dev);
> +       if (!mvi) {
> +               res = -1;
> +               goto found_out;
> +       }
>
>         if (lock)
>                 spin_lock_irqsave(&mvi->lock, flags);
> --
> 2.5.0
>

It doesn't look right,
if mvi will be NULL and lock will be set then at

found_out:
    if (lock)
        spin_unlock_irqrestore(&mvi->lock, flags);

there will be mvi dereference, besides spin_lock_irqsave wasn't even called.
And without this patch dereference would happen on mvi->lock which is
before use in mvs_alloc_dev

About whether mvs_find_dev_mvi can return NULL it looks like it's possible,
but I'm not sure if it practically happens. I guess it did hence patch.