Miouyouyou
commented
5 years ago Enabling DRM_FBDEV_LEAK_PHYS_SMEM changes nothing...
Closed Miouyouyou closed 4 years ago
Miouyouyou
commented
5 years ago Enabling DRM_FBDEV_LEAK_PHYS_SMEM changes nothing...
Miouyouyou
commented
5 years ago The new "not-working" set of patches applied to v4.20 kernels is available here :
https://github.com/Miouyouyou/RockMyy/tree/v4.20-IsNotWorkingToo
Tonymac32
commented
5 years ago Hmmm, there were a lot of notes about the phy clocks on the dwmmc rockchip, prior to these patches you got the errors while it parsed through invalid clock rates. Related maybe?
Miouyouyou
commented
5 years ago Ah, you mean the patches provided by the ASUS team on their 4.4 ? Or the invalid Clock rates in the dmesg ?
Come to say, that reminds me an old issue with eMMC on MiQi devices. I guess I could try to boot a MiQi with a 4.20 and see how it fares for the MMC.
Still, my biggest issue is the Mali kernel driver failing its mmap calls... It just break the entire OpenGL stack and X11 one too.
Tonymac32
commented
5 years ago Right, let me poke around on the mmc stuff, I might be confusing some rk3328 stuff in the mix at the same time. Also, Tinker device tree patches will be a mess eventually (if not already) because they made a tinker.dtsi with individual dts's for the Tinker and "S"
Miouyouyou
commented
5 years ago Yeah... The DTSI thing... I know...
... I had to rewrite like ~10 patches, one by one, due to these modifications.
I like how the tinker.dts content is now like "I'm a Tinkerboard. The End."
Tonymac32
commented
5 years ago Yep...
Tonymac32
commented
5 years ago I noticed a SERDEV situation on hci-3wire got backported to 4.4, I need to check that out as I can't seem to follow the bluetooth discussion properly
Tonymac32
commented
5 years ago Miouyouyou
commented
5 years ago I've been able to "kind of" fix the Mali issues, but it's a more a kludge than anything, and some problems still remain.
It seems that the switch to vmf_insert_pfn was done through multiple Linux versions. Now, the NVIDIA drivers seem to be impacted by the same bug, so I'll take a look at other kernel drivers that still use old ABI and see how people fixed their issue with these drivers.
Meanwhile, the ALSA drivers are now dropping this during the boot sequence :
[ 12.550485] ------------[ cut here ]------------
[ 12.555656] kernel BUG at ./include/linux/dma-mapping.h:327!
[ 12.561983] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM
[ 12.564945] dwmmc_rockchip ff0c0000.dwmmc: Unexpected interrupt latency
[ 12.568505] Modules linked in: r8723bs(C) mali_kbase snd_soc_simple_card snd_soc_simple_card_utils rockchip_rga dw_hdmi_cec v4l2_mem2mem videobuf2_dma_sg rk_crypto
[ 12.592223] CPU: 0 PID: 1598 Comm: alsa-sink-ff890 Tainted: G C 4.20.0-rc2-RockMyyX2 #1
[ 12.602627] Hardware name: Rockchip (Device Tree)
[ 12.607887] PC is at pl330_prep_slave_fifo+0x78/0x10c
[ 12.613530] LR is at 0xc145e564
[ 12.617033] pc : [<c07012bc>] lr : [<c145e564>] psr: 20010093
[ 12.624032] sp : eadc5e40 ip : 40000000 fp : 0202f700
[ 12.629864] r10: 00000001 r9 : 00000000 r8 : c0e0144c
[ 12.635696] r7 : 00000002 r6 : 00000000 r5 : ef033800 r4 : eeac0000
[ 12.642985] r3 : c145e564 r2 : 0000000c r1 : 00000000 r0 : 00000001
[ 12.650276] Flags: nzCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment none
[ 12.651927] dwmmc_rockchip ff0d0000.dwmmc: Unexpected interrupt latency
[ 12.658335] Control: 10c5387d Table: 2ac9006a DAC: 00000051
[ 12.672134] Process alsa-sink-ff890 (pid: 1598, stack limit = 0x2c06a11e)
[ 12.679716] Stack: (0xeadc5e40 to 0xeadc6000)
[ 12.684580] 5e40: eeac0000 00000001 00001138 00000000 000044e0 eeac0000 eea66c18 00001138
[ 12.693717] 5e60: 2cd38000 c0702898 00001138 00000000 eac3a7c0 ef23b600 c070282c 00000003
[ 12.702853] 5e80: ee0f3a00 00000036 0202f700 c0a48d24 00000001 00000003 ed4b7c40 edece800
[ 12.711981] 5ea0: ef23b600 00000001 edecedcc c0a74db4 ef23b600 00000003 c0ee5728 00000000
[ 12.721118] 5ec0: eac3a640 00000023 00000036 c0a4049c 00004142 02050878 ef23b600 eac25d00
[ 12.730254] 5ee0: eac3a640 c0a430e0 006200ca 00000004 b5d0c000 eac92d70 eac92d70 2cd3c707
[ 12.739390] 5f00: 00000000 02050878 ecd728e8 eac25d00 00004142 02050878 00000023 c02421dc
[ 12.748518] 5f20: 02050878 c0242900 00000055 ed7dd780 0000080f ad0fc950 ece15200 c01132b0
[ 12.757655] 5f40: c130a488 00000000 00000001 00000000 ece15248 00000800 00000000 000000f0
[ 12.766792] 5f60: eac25d00 c024c938 00000000 eac25d00 eac25d01 00004142 02050878 00000023
[ 12.775928] 5f80: 00000036 c02430e8 020511f0 adb4d000 00000000 00000036 c0101204 eadc4000
[ 12.785065] 5fa0: 00000036 c0101000 020511f0 adb4d000 00000023 00004142 02050878 02050878
[ 12.794201] 5fc0: 020511f0 adb4d000 00000000 00000036 00000000 00000001 adaa1000 0202f700
[ 12.803337] 5fe0: adb4d100 ad0feca4 adae1df3 b6cbae06 20010030 00000023 00000000 00000000
[ 12.812480] [<c07012bc>] (pl330_prep_slave_fifo) from [<c0702898>] (pl330_prep_dma_cyclic+0x6c/0x1f0)
[ 12.822788] [<c0702898>] (pl330_prep_dma_cyclic) from [<c0a48d24>] (snd_dmaengine_pcm_trigger+0xc4/0x170)
[ 12.833485] [<c0a48d24>] (snd_dmaengine_pcm_trigger) from [<c0a74db4>] (soc_pcm_trigger+0xec/0x128)
[ 12.843597] [<c0a74db4>] (soc_pcm_trigger) from [<c0a4049c>] (snd_pcm_action_single+0x38/0x64)
[ 12.853221] [<c0a4049c>] (snd_pcm_action_single) from [<c0a430e0>] (snd_pcm_ioctl+0x824/0xf6c)
[ 12.862848] [<c0a430e0>] (snd_pcm_ioctl) from [<c02421dc>] (vfs_ioctl+0x28/0x3c)
[ 12.871114] [<c02421dc>] (vfs_ioctl) from [<c0242900>] (do_vfs_ioctl+0x84/0x81c)
[ 12.879377] [<c0242900>] (do_vfs_ioctl) from [<c02430e8>] (ksys_ioctl+0x50/0x6c)
[ 12.887642] [<c02430e8>] (ksys_ioctl) from [<c0101000>] (ret_fast_syscall+0x0/0x54)
[ 12.896194] Exception stack(0xeadc5fa8 to 0xeadc5ff0)
[ 12.901834] 5fa0: 020511f0 adb4d000 00000023 00004142 02050878 02050878
[ 12.910970] 5fc0: 020511f0 adb4d000 00000000 00000036 00000000 00000001 adaa1000 0202f700
[ 12.920095] 5fe0: adb4d100 ad0feca4 adae1df3 b6cbae06
[ 12.925738] Code: e1a00626 ebe84905 e3500000 0a000001 (e7f001f2)
[ 12.932543] ---[ end trace 16d29d6dead4fbbb ]---These issues are not relevant anymore. Closing up.
It seems that the new patches applied to kernel 4.20-rc1 generate issues with the Mali Kernel Driver and the dwmmc.
While the Mali Blob might have some issues with a new "deprecated" option added to the Framebuffer driver, that is not enabled in the new configuration, the DWMMC seems very weird.