Hi Janusz, On 2026-02-05 at 15:51:55 +0100, Janusz Krzysztofik wrote: > In a short listing, lsgpu prints a sysfs path of a PCI GPU parent as a > local attribute of a DRM device. However, if that's a discrete GPU and > its associated PCIe upstream bridge port has been identified, no > information on that bridge is listed among the GPU attributes. Follow the > pattern used with DRM devices and also show a PCI slot of the bridge port > as a local attribute of the discrete GPU device. > > Moreover, in both short and detailed listings, local attributes intended > for providing device names of GPU associated DRM devices and the GPU > codename are also printed as attributes of related PCIe upstream bridge > port, however, the DRM device names are shown as (null), and the codename > attribute provides raw vendor:device codes of the bridge itself. Replace > those with PCI slot and codename of the GPU device. > > v4: Free already allocated numeric codename before replacing (Sebastian), > - visually separate bridge specific updates of attributes already > populated by find_or_add_igt_device from bridge specific attribute > assignments. > v2: Allocate memory to local attributes of a bridge for safety (Sebastian), > - merge with a formerly separate patch "lib/igt_device_scan: Don't print > bridge not applicable attributes" (Sebastian), > - no need for DEVTYPE_BRIDGE, just skip attributes if NULL. > > Cc: Sebastian Brzezinka <[email protected]> > Signed-off-by: Janusz Krzysztofik <[email protected]> > ---
Reviewed-by: Krzysztof Karas <[email protected]> -- Best Regards, Krzysztof
