re: kern/54976: radeon framebuffer fails initialization in NetBSD-9

[matthew green <mrg%eterna.com.au@localhost>]

The following reply was made to PR kern/54976; it has been noted by GNATS.

From: matthew green <mrg%eterna.com.au@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
    netbsd-bugs%netbsd.org@localhost, rcbixler%nyx.net@localhost
Subject: re: kern/54976: radeon framebuffer fails initialization in NetBSD-9
Date: Wed, 19 Feb 2020 17:12:06 +1100

 OK, i think i see the real problem.
 
 quick survey of the request_firmware() calls in radeon
 shows that most attempt both names, eg:
 
         case CHIP_KAVERI:
                 chip_name =3D "KAVERI";
                 new_chip_name =3D "kaveri";
 ...
         snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", new_chip_n=
 ame);
         err =3D request_firmware(&rdev->pfp_fw, fw_name, rdev->dev);
         if (err) {
                 snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", ch=
 ip_name);
                 err =3D request_firmware(&rdev->pfp_fw, fw_name, rdev->dev=
 );
 ...
 
 but the KAVERI path doesn't look for old chip_name:
 
         if (rdev->family =3D=3D CHIP_KAVERI) {
                 snprintf(fw_name, sizeof(fw_name), "radeon/%s_mec2.bin", n=
 ew_chip_name);
                 err =3D request_firmware(&rdev->mec2_fw, fw_name, rdev->de=
 v);
                 if (err) {
                         goto out;
 
 i guess this just needs to also check both names.
 
 
 .mrg.