linux-kernel - Re: [PATCH 04/22] firmware: arm_scmi: Validate BASE_DISCOVER_LIST

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: <YmqaSZJlPF2qX5Ta@e120937-lin>
Date:   Thu, 28 Apr 2022 14:45:07 +0100
From:   Cristian Marussi <cristian.marussi@....com>
To:     Sudeep Holla <sudeep.holla@....com>
Cc:     linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        james.quinlan@...adcom.com, Jonathan.Cameron@...wei.com,
        f.fainelli@...il.com, etienne.carriere@...aro.org,
        vincent.guittot@...aro.org, souvik.chakravarty@....com
Subject: Re: [PATCH 04/22] firmware: arm_scmi: Validate
 BASE_DISCOVER_LIST_PROTOCOLS reply

On Thu, Apr 28, 2022 at 11:07:29AM +0100, Sudeep Holla wrote:
> On Wed, Mar 30, 2022 at 04:05:33PM +0100, Cristian Marussi wrote:
> > Do not blindly trust SCMI backend server reply about list of implemented
> > protocols, instead validate the reported length of the list of protocols
> > against the real payload size of the message reply.
> >
> > Fixes: b6f20ff8bd9 ("firmware: arm_scmi: add common infrastructure and support for base protocol")
> > Signed-off-by: Cristian Marussi <cristian.marussi@....com>
> > ---
> >  drivers/firmware/arm_scmi/base.c | 21 +++++++++++++++++++++
> >  1 file changed, 21 insertions(+)
> >
> > diff --git a/drivers/firmware/arm_scmi/base.c b/drivers/firmware/arm_scmi/base.c
> > index f279146f8110..c1165d1282ef 100644
> > --- a/drivers/firmware/arm_scmi/base.c
> > +++ b/drivers/firmware/arm_scmi/base.c
> > @@ -189,6 +189,9 @@ scmi_base_implementation_list_get(const struct scmi_protocol_handle *ph,
> >  	list = t->rx.buf + sizeof(*num_ret);
> >
> >  	do {
> > +		size_t real_list_sz;
> > +		u32 calc_list_sz;
> > +
> >  		/* Set the number of protocols to be skipped/already read */
> >  		*num_skip = cpu_to_le32(tot_num_ret);
> >
> > @@ -202,6 +205,24 @@ scmi_base_implementation_list_get(const struct scmi_protocol_handle *ph,
> >  			break;
> >  		}
> >
> > +		if (t->rx.len < (sizeof(u32) * 2)) {
> > +			dev_err(dev, "Truncated reply - rx.len:%zd\n",
> > +				t->rx.len);
> > +			ret = -EPROTO;
> > +			break;
> > +		}
> > +
> > +		real_list_sz = t->rx.len - sizeof(u32);
> > +		calc_list_sz = ((loop_num_ret / sizeof(u32)) +
> > +				!!(loop_num_ret % sizeof(u32))) * sizeof(u32);
> 
> Any reason this can't be (loop_num_ret - 1) / sizeof(u32) + 1 ?
> 

At first sight could be fine with your easier version BUT what if loop_num_ret
is returned as zero ?

real_list_sz should be ZERO length and calc_list_sz

im my version:

calc_list_sz = ((0/4) +!!(0%4)) * 4   ===>> 0

while in the simplified one gets calculated wrong:

calc_list_sz = (0-1)/4 + 1 ====> 1

...moreover being both loop_num_ret and calc_list_sz unsigned I am even
not so sure about implicit casting messing things up evenm more :D

So I sticked to the more convoluted approach :D

....Have I missed something else ?

Thanks,
Cristian