linux-kernel - Re: [PATCH] spi: Raise limit on number of chip selects

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]

Date: Tue, 23 Jan 2024 09:47:17 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: Jonas Gorski <jonas.gorski@...il.com>, Mark Brown <broonie@...nel.org>
Cc: Christophe Leroy <christophe.leroy@...roup.eu>,
 Herve Codina <herve.codina@...tlin.com>,
 Amit Kumar Mahapatra <amit.kumar-mahapatra@....com>,
 "linux-spi@...r.kernel.org" <linux-spi@...r.kernel.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
 Thomas Petazzoni <thomas.petazzoni@...tlin.com>
Subject: Re: [PATCH] spi: Raise limit on number of chip selects

On 1/23/24 08:50, Jonas Gorski wrote:
> Hi,
> 
> On Tue, 23 Jan 2024 at 14:56, Mark Brown <broonie@...nel.org> wrote:
>>
>> On Tue, Jan 23, 2024 at 01:26:04PM +0000, Christophe Leroy wrote:
>>> Le 23/01/2024 à 14:18, Mark Brown a écrit :
>>>> On Tue, Jan 23, 2024 at 12:04:30PM +0100, Herve Codina wrote:
>>
>>>>> Moving the SPI_CS_CNT_MAX value from 4 to 8 is not enough to handle my case.
>>>>> Tested moving SPI_CS_CNT_MAX to 16 and it was ok.
>>
>>>> OK, I've also heard 12 as a number which this would cover.
>>
>>> By the way the comment in include/linux/spi/spi.h is confusing. This
>>> SPI_CS_CNT_MAX is really not the max number of CS supported per SPI
>>> device but the max number of CS supported per SPI controller.
>>
>> Well, it's a combination of the comment being confusing and the
>> implementation being a bit broken - we simply shouldn't be limiting the
>> number of chip selects per controller, the per device limit is much more
>> reasonable.  So ideally the code would be changed to reflect the
>> comment.
> 
> At a first glance at all places using SPI_CS_CNT_MAX I don't see
> anything being broken / reading out of bounds if a controller has more
> chipselects than SPI_CS_CNT_MAX.
> 
> So I think the check of ctrl->num_chipselect in of_spi_parse_dt() is
> bogus/unnecessary and is in the wrong place, as this is for parsing a
> spi device node and not a controller node. The following check for the
> amount of chip selects defined for the spi device should just check
> against SPI_CS_CNT_MAX instead of ctrl->num_chipselects.
> __spi_add_device() later will ensure that any chip selects are valid
> chip selects, so no need for of_spi_parse_dt() to check that either.
> 
> But I didn't do a very thorough read, or even tested it, so I might
> have easily missed something.
> 

struct spi_controller {
	...
	char                            last_cs[SPI_CS_CNT_MAX];

does introduce the limit for controllers.

Guenter