linux-kernel - Re: [PATCH 3/7] mtd: spi-nor: add restriction for nmaps in smpt parser

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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Thu, 8 Nov 2018 15:15:09 +0100
From:   Boris Brezillon <boris.brezillon@...tlin.com>
To:     <Tudor.Ambarus@...rochip.com>
Cc:     <marek.vasut@...il.com>, <dwmw2@...radead.org>,
        <computersforpeace@...il.com>, <richard@....at>,
        <linux-mtd@...ts.infradead.org>, <linux-kernel@...r.kernel.org>,
        <yogeshnarayan.gaur@....com>, <cyrille.pitchen@...ev4u.fr>
Subject: Re: [PATCH 3/7] mtd: spi-nor: add restriction for nmaps in smpt
 parser

On Thu, 8 Nov 2018 13:58:45 +0000
<Tudor.Ambarus@...rochip.com> wrote:

> On 11/08/2018 02:54 PM, Boris Brezillon wrote:
> > On Thu, 8 Nov 2018 11:07:11 +0000
> > <Tudor.Ambarus@...rochip.com> wrote:
> >   
> >> The map selector is limited to a maximum of 8 bits, allowing
> >> for a maximum of 256 possible map configurations. The total
> >> number of map configurations should be addressable by the
> >> total number of bits described by the detection commands.
> >>
> >> For example: if there are five to eight possible sector map
> >> configurations, at least three configuration detection commands
> >> will be needed to extract three bits of configuration selection
> >> information from the device in order to identify which configuration
> >> is currently in use.
> >>
> >> Suggested-by: Boris Brezillon <boris.brezillon@...tlin.com>  
> > 
> > I don't remember suggesting this :-).  
> 
> I see this when you discussed with Yogesh:
> 
> +       for (nmaps = 0; i< smpt_len; nmaps++) {
> +               if(!(smpt[i] & SMPT_DESC_TYPE_MAP)) {
> +                       i += 2;
> +                       continue;
> +               }
> +
> +               if(!map_id_is_valid) {
> +                       if (nmaps) {
> +                               ret = NULL;
> +                               break;
> +                       }
> 
> If I understand correctly, you meant that if there are no detection commands,
> and more than a configuration map, then return an error.

Yes, because in this case we have no way to know what config is
currently selected.

> 
> This is correct in my opinion. What I did was to generalize this idea. If smtp
> defines more maps than you can select using detection commands, return error.

AFAICT you're no longer checking that map_id is valid (see below).

> 
> +
> +                       ret = smpt+i;
> +               } else if (map_id == SMPT_MAP_ID(smpt[i])) {
> +                       ret = smpt+i;
> +                       break;
> +               }
> +
>                 /* increment the table index to the next map */
>                 i += SMPT_MAP_REGION_COUNT(smpt[i]) + 1;
>         }
> 
> >   
> >> Signed-off-by: Tudor Ambarus <tudor.ambarus@...rochip.com>
> >> ---
> >>  drivers/mtd/spi-nor/spi-nor.c | 15 +++++++++++++--
> >>  1 file changed, 13 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/mtd/spi-nor/spi-nor.c b/drivers/mtd/spi-nor/spi-nor.c
> >> index 59dcedb08691..bd1866d714f2 100644
> >> --- a/drivers/mtd/spi-nor/spi-nor.c
> >> +++ b/drivers/mtd/spi-nor/spi-nor.c
> >> @@ -2868,7 +2868,7 @@ static const u32 *spi_nor_get_map_in_use(struct spi_nor *nor, const u32 *smpt,
> >>  	const u32 *ret = NULL;
> >>  	u32 addr;
> >>  	int err;
> >> -	u8 i;
> >> +	u8 i, ncmds, nmaps;
> >>  	u8 addr_width, read_opcode, read_dummy;
> >>  	u8 read_data_mask, data_byte, map_id;
> >>  
> >> @@ -2877,6 +2877,7 @@ static const u32 *spi_nor_get_map_in_use(struct spi_nor *nor, const u32 *smpt,
> >>  	read_opcode = nor->read_opcode;
> >>  
> >>  	map_id = 0;
> >> +	ncmds = 0;
> >>  	/* Determine if there are any optional Detection Command Descriptors */
> >>  	for (i = 0; i < smpt_len; i += 2) {
> >>  		if (smpt[i] & SMPT_DESC_TYPE_MAP)
> >> @@ -2896,6 +2897,7 @@ static const u32 *spi_nor_get_map_in_use(struct spi_nor *nor, const u32 *smpt,
> >>  		 * Configuration that is currently in use.
> >>  		 */
> >>  		map_id = map_id << 1 | !!(data_byte & read_data_mask);
> >> +		ncmds++;
> >>  	}
> >>  
> >>  	/*
> >> @@ -2905,7 +2907,16 @@ static const u32 *spi_nor_get_map_in_use(struct spi_nor *nor, const u32 *smpt,
> >>  	 *
> >>  	 * Find the matching configuration map.
> >>  	 */
> >> -	while (i < smpt_len) {
> >> +	for (nmaps = 0; i < smpt_len; nmaps++) {
> >> +		/*
> >> +		 * The map selector is limited to a maximum of 8 bits, allowing
> >> +		 * for a maximum of 256 possible map configurations. The total
> >> +		 * number of map configurations should be addressable by the
> >> +		 * total number of bits described by the detection commands.
> >> +		 */
> >> +		if (ncmds && nmaps >= (1 << (ncmds + 1)))
> >> +			break;

You're no longer checking that when ncmds == 0 nmaps has to be 1. So,
say the chip exposed no smpt commands and has several maps defined,
map_id will be 0 while it should have be "undefined". My version
would return an error, but yours tries to find map_id 0.

> >> +  
> > 
> > Maybe I missed something but it sounds like this change is just
> > optimizing the SPMT parsing a bit, and to be honest, I'm not sure this
> > is really needed. Most of the time, smpt_len will be rather small, so
> > trying to bail out earlier is not bringing much perf improvements.  
> 
> It's rather a smtp validity check. I want to return an error if there are not
> enough detection commands to identify the map id.

You would have failed the same way without this validity check after a
maximum of smpt_len iterations, right?