Hi,

On 2025-05-14 15:39, Michael Lettrich wrote:
> Package: libsensors5
> Version: 1:3.6.0-7.1 arm64
> Severity: normal
> Tags: patch upstream
> X-Debbugs-Cc: michael.lettr...@cern.ch, federico.v...@cern.ch
> 
> Dear Maintainer,
> 
> we have found an issue in libsensors that occurs when parsing the chip
> address of ISA sensors resulting in an incorrect ISA address.
> 
> The ISA sensor chip address of a sensor is obtained by parsing the
> "device" symlink of an hwmon device. The regular expression extracting
> the device address will fail if the "device" symlink contains dashes
> and sets the address to 0. If multiple instances of the same board are
> plugged into the same system, cards can no longer be distinguished, as
> they all will  be incorrectly be assigned an address of 0.
> 
> Example: 
> const char* dev_name = "ucsi-source-psy-USBC000:001";
> int addr = -1;
> /* current regex */
> if (sscanf(dev_name, "%*[a-zA-Z0-9_]%*1[.:]%d", &addr) != 1)
>    addr= 0;
> printf("addr: %d\n", addr); /* prints 0 */
> addr = -1;
> /* proposed regex */
> if (sscanf(dev_name, "%*[a-zA-Z0-9_-]%*1[.:]%d", &addr) != 1)
>    addr= 0;
> printf("addr: %d\n", addr); /* prints 1 */
> 
> 
> We propose to add dashes to the regular expression, so that device
> names with dashes will be correctly recognized, and provide it as a
> patch. Note that we already opened an issue upstream at 
> https://github.com/lm-sensors/lm-sensors/pull/529
> It seems though, that the package is abandoned for a few years now. It
> would be great if we could at least fix things in Debian.

The original lm-sensors upstream is indeed dead. Debian and a few other 
distributions (Arch, SUSE, Gentoo, ...) have switched to a new upstream: 
https://github.com/hramrach/lm-sensors . You might want to submit your 
patch there (note that there have been changes the few lines around).

If it doesn't get merged there, we can indeed take it as a Debian 
specific change.

Regards
Aurelien

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurel...@aurel32.net                     http://aurel32.net

Reply via email to