<kwizart>
digetx, yes, if the governor is available when tegra20-emc is loaded (like within an initramfs) it will auto-load it, but for both to be available in initramfs, there is a need to add a certain tag in the tegra20-emc (like: MODULE_SOFTDEP("pre: governor_simpleondemand")
<digetx>
kwizart: how initramfs module list is created in the first place? are you specifying modules manually?
<kwizart>
digetx, well, I can patch dracut to do the right thing, but modules list are created according to either it's a disk driver or a dependency of a disk driver (usb/ pci/ etc) or a graphic driver. Opionally graphic drivers can be added for desktop users
<kwizart>
so either tegra20-emc is relevant for initramfs is a question...
<kwizart>
currently, tegra20-emc might not be added by default, I have added it explicitely, but I can double check if it's done by default (or if not having it leads to issue)
e1z0 has quit [Read error: Connection reset by peer]
e1z0 has joined #tegra
e1z0 has quit [Read error: Connection reset by peer]
e1z0 has joined #tegra
<digetx>
kwizart: drm won't work without tegra20-emc
<digetx>
softdep will work for emc
<digetx>
but still drm driver needs interconnect providers, i.e. memory drivers
<digetx>
it's possible to list all memory drivers as softdeps for drm driver, but then drm will load unrelated memory modules unconditionally
<kwizart>
digetx, maybe it's possible to use: (like: MODULE_SOFTDEP("post: tegra_drm") from tegra20-emc... so tegra_drm will not have always have it if tegra20-emc isn't relevant for the given hardward
<digetx>
that won't help drm to express dependency on the emc
<kwizart>
no, but that will help the emc to express that it's needed for tegra_drm
<kwizart>
I haven't tested much that way to express "reverse dependencies" but that's testable
<digetx>
need a way to specify which modules should present in ramfs without enforcing preloading of those modules
<digetx>
maybe dracut just needs option for "memory" drivers
<kwizart>
another way would be to tell that all memory module needs to be available
<kwizart>
indeed, we are using them builtin in Fedora, but they can be built as modules, indeed
<kwizart>
or maybe I should double check
<kwizart>
yes fedora has CONFIG_TEGRA_IOMMU_{GART,SMMU}=y
<kwizart>
and tegra_drm has depends: host1x,drm,drm_kms_helper,cec,iova
<digetx>
tegra iommu drivers can't be built as modules, but there is arm smmu and other drivers that could
<digetx>
on the other hand, I'm not sure whether those drivers could really be useful in practice when built as modules
<kwizart>
which ones ?
<digetx>
all iommu drivers
<kwizart>
interesting is that despite been listed as depends: iova isn't automatically added into the initramfs, maybe thats the error I'm seeing with 5.14:
<kwizart>
Failed to attached device 54200000.dc to IOMMU_mapping (on jetson-tk1)
<digetx>
drm driver won't load at all without iova
<kwizart>
so that's probably a reason why tegra_drm is taking too long to load
<kwizart>
is iova needed also in tegra20 ?
<digetx>
it's needed unconditionally for all tegras since tegra-drm uses symbols from the iova module
<digetx>
"Failed to attached device" should be harmless
<digetx>
you should have it on older kernels that work
<digetx>
dracut should automatically pull iova and others modules by the used symbols, are you sure that CONFIG_IOVA=m?
<kwizart>
okay iova is added as a dependency of the tegra_drm only for cases when the drm (optional) module is used by dracut as I understand
<digetx>
that's correct
<digetx>
for tegra_drm you need memory,regulator,gpio modules
<digetx>
and host1x
<digetx>
but host1x should be auto-pulled
marvin24 has quit [Ping timeout: 252 seconds]
marvin24 has joined #tegra
torez has joined #tegra
torez has quit [Ping timeout: 245 seconds]
torez has joined #tegra
marvin24 has quit [Ping timeout: 252 seconds]
marvin24_ has joined #tegra
MatrixTravelerbo has quit [Quit: Bridge terminating on SIGTERM]
AntoniAloyTorren has quit [Quit: Bridge terminating on SIGTERM]
ServerStatsDisco has quit [Quit: Bridge terminating on SIGTERM]
DavidHeidelberg has quit [Quit: Bridge terminating on SIGTERM]
x86corez has quit [Quit: Bridge terminating on SIGTERM]
thevengefulprinc has quit [Quit: Bridge terminating on SIGTERM]
nergzd723[m] has quit [Quit: Bridge terminating on SIGTERM]
gavodavo has quit [Quit: Bridge terminating on SIGTERM]
jenneron[m] has quit [Quit: Bridge terminating on SIGTERM]
Jasper[m] has quit [Quit: Bridge terminating on SIGTERM]
wolfreiser[m] has quit [Quit: Bridge terminating on SIGTERM]
kusma has quit [Quit: Bridge terminating on SIGTERM]
aonbehamut[m] has quit [Quit: Bridge terminating on SIGTERM]
jonasschwoebel[m has quit [Quit: Bridge terminating on SIGTERM]
pgwipeout[m] has quit [Quit: Bridge terminating on SIGTERM]
leanderglanda[m] has quit [Quit: Bridge terminating on SIGTERM]
osan[m] has quit [Quit: Bridge terminating on SIGTERM]
pangelo[m] has quit [Quit: Bridge terminating on SIGTERM]
maxnet[m] has quit [Quit: Bridge terminating on SIGTERM]
maxim[m] has quit [Quit: Bridge terminating on SIGTERM]
gouchi has quit [Remote host closed the connection]
gouchi has joined #tegra
digetx has joined #tegra
<m4t>
digetx: just got same panic again, rebuilding with kasan (CONFIG_KASAN_INLINE=y)
<m4t>
i think it's happening when munin runs 'sensors' cmd... not 100% sure, but looking at .bash_history when i ping/ssh when it freezes, seems to line up
<m4t>
'ouya' is alias to ssh in: 317399 10/11/21 22:40:27 ouya | 317538 10/14/21 16:05:26 ouya