DavidHeidelberg has quit [*.net *.split]
DavidHeidelberg has joined #tegra
DavidHeidelberg has quit [Ping timeout: 245 seconds]
DavidHeidelberg has joined #tegra
digetx has quit [Ping timeout: 264 seconds]
digetx has joined #tegra
danieel has quit [Ping timeout: 272 seconds]
danieel has joined #tegra
danieel has quit [Ping timeout: 268 seconds]
danieel has joined #tegra
<
cyndis>
yeah, on T210, we're very close to feature complete
<
cyndis>
NVENC and NVJPG need some userspace written, camera support is in staging as well
kwizart has joined #tegra
digetx_ has joined #tegra
digetx has quit [Killed (NickServ (GHOST command used by digetx_))]
<
cyndis>
v6 now out there
<
cyndis>
userspaces changes too for grate, opentegra, and vaapi driver (including the nvdec support)
<
cyndis>
i mean h264 support
<
digetx_>
cyndis: super, thank you
<
digetx_>
i"m starting to test it
<
digetx_>
ERROR: drm_channel_submit2:1023: ioctl(DRM_IOCTL_TEGRA_CHANNEL_SUBMIT) failed: 22
<
digetx_>
ERROR: grate_overlay_show: host1x_overlay_set() failed: -22
<
cyndis>
weird, I tested cube-textured2 just before sending and that was working fine
<
cyndis>
yeah, working fine here
<
cyndis>
I'm on a -next a couple months old though
<
cyndis>
don't know how that would affect the firewall though
<
digetx_>
that is with the v6 on top of recent next, trying your kernel branch now
<
digetx_>
cyndis: your branch a bit better, there is no kernel firewall error message and at least 3d seem to work, but 2d not
<
digetx_>
ERROR: drm_channel_submit2:1023: ioctl(DRM_IOCTL_TEGRA_CHANNEL_SUBMIT) failed: 22
<
digetx_>
INFO: grate_texture_generate_mipmap: LOD 2 w: 64 h: 64 offset 0x00040000 -> 0x00050000
<
digetx_>
ERROR: drm_channel_submit2:1023: ioctl(DRM_IOCTL_TEGRA_CHANNEL_SUBMIT) failed: 22
<
digetx_>
INFO: grate_texture_generate_mipmap: LOD 1 w: 128 h: 128 offset 0x00000000 -> 0x00040000
<
digetx_>
ERROR: drm_channel_submit2:1023: ioctl(DRM_IOCTL_TEGRA_CHANNEL_SUBMIT) failed: 22
<
digetx_>
INFO: grate_texture_generate_mipmap: LOD 3 w: 32 h: 32 offset 0x00050000 -> 0x00054000
<
cyndis>
do you know of any potential difference in t20/t30 vs. t114 that could cause it? and is there any kernel log message?
<
digetx_>
cube-textured2 works, texture-filter not
<
cyndis>
texture-filter is working for me
<
digetx_>
there shouldn't be any differences for t114, kernel log is silent
digetx_ is now known as digetx
<
cyndis>
interesting, I added prints on pretty much every error path during submit
<
cyndis>
only places where it's missing is on host1x_job_pin and host1x_fence_create but grate should not use fences
<
cyndis>
maybe you have CONFIG_*_FIREWALL enabled, and it's failing the job since it's not getting the reloc information
<
cyndis>
might need to temporarily add a way to disable the old firewall when coming through the new uapi
<
digetx>
will push the fix or should I take a look?
<
cyndis>
please try to disable CONFIG_*_FIREWALL
<
cyndis>
if it works I'll add a fix for that when pushing v7
<
digetx>
will be also nice if you could rebase on the recent next
<
cyndis>
yeah, will do that as well
<
cyndis>
but if it's working with my branch and can be tested, I'll do it over next week, already spent most of the weekend on this
<
cyndis>
can you send your .config file
<
digetx>
I'm trying it on T30, will check T20 now
<
digetx>
tegra-gr3d 54180000.gr3d: illegal INCR(offset=0x102, count=2) in class 0x60 at word 1419 on T20
<
cyndis>
ok, i'll try to reproduce with your config, will debug more tomorrow
<
digetx>
same illegal INCR on T30 with CONFIG_TEGRA_IOMMU_SMMU=n
<
digetx>
the mapping->sgt should be NULL without IOMMU, not sure why channel_map doesn't fail
<
digetx>
ah no, it shouldn't
<
digetx>
but the mapping logic looks broken anyways
<
digetx>
actually, it might be broken in general, I'll check the vanilla upstream now
<
digetx>
vanilla upstream is broken
<
digetx>
but that breakage is unrelated to mapping, PIX_BUF_LAYOUT_TILED_16x16 is broken
digetx has quit [Ping timeout: 268 seconds]
digetx has joined #tegra
Guest6 has joined #tegra
Guest6 has quit [Client Quit]