| Message ID | 20180712154128.22705-6-thierry.reding@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [GIT,PULL,1/5] dt-bindings: tegra: Changes for v4.19-rc1 | expand |
On Thu, Jul 12, 2018 at 05:41:28PM +0200, Thierry Reding wrote: > Hi ARM SoC maintainers, > > The following changes since commit ce397d215ccd07b8ae3f71db689aedb85d56ab40: > > Linux 4.18-rc1 (2018-06-17 08:04:49 +0900) > > are available in the Git repository at: > > git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux.git tags/tegra-for-4.19-arm64-dt > > for you to fetch changes up to 7780a03495e13cd2bef704bcbf8c727de9f65232: > > arm64: tegra: Add CPU nodes to Tegra194 device tree (2018-07-02 15:57:39 +0200) > > Thanks, > Thierry > > ---------------------------------------------------------------- > arm64: tegra: Device tree changes for v4.19-rc1 > > These changes enable the GPIO controllers on Tegra194 SoCs, which in > turn allows the SD card detection and ethernet controllers to be enabled > as well. The Tegra194 device tree is also extended with the list of CPUs > and a PSCI node to inform the kernel about the presence of PSCI capable > firmware. Merged, thanks. Are you planning on supporting 194 better than 186? That never really made it to a useful level upstream I think? :( -Olof
On Sat, Jul 14, 2018 at 02:22:10PM -0700, Olof Johansson wrote: > On Thu, Jul 12, 2018 at 05:41:28PM +0200, Thierry Reding wrote: > > Hi ARM SoC maintainers, > > > > The following changes since commit ce397d215ccd07b8ae3f71db689aedb85d56ab40: > > > > Linux 4.18-rc1 (2018-06-17 08:04:49 +0900) > > > > are available in the Git repository at: > > > > git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux.git tags/tegra-for-4.19-arm64-dt > > > > for you to fetch changes up to 7780a03495e13cd2bef704bcbf8c727de9f65232: > > > > arm64: tegra: Add CPU nodes to Tegra194 device tree (2018-07-02 15:57:39 +0200) > > > > Thanks, > > Thierry > > > > ---------------------------------------------------------------- > > arm64: tegra: Device tree changes for v4.19-rc1 > > > > These changes enable the GPIO controllers on Tegra194 SoCs, which in > > turn allows the SD card detection and ethernet controllers to be enabled > > as well. The Tegra194 device tree is also extended with the list of CPUs > > and a PSCI node to inform the kernel about the presence of PSCI capable > > firmware. > > Merged, thanks. Are you planning on supporting 194 better than 186? That never > really made it to a useful level upstream I think? :( I think we have pretty decent upstream support for Tegra186 at this point. It took a little time because Tegra186 was a little work than many previous chips because the changes were less incremental than before. Generally our goal is to provide at least feature parity with prior chips for each new chip. Usually we manage to incrementally add new features, too. Perhaps the reason why it seems that Tegra186 is not well supported upstream is because a lot of things have moved into BPMP firmware, which deals with a lot of the nitty gritty details that prior chips needed a driver for. That said, I may be missing something. Do you have any specific features in mind that aren't working upstream on Tegra186? Tegra194 uses the same firmware as Tegra186 (well, a binary compatible interface to the firmware, I should say), so we've been able to bootstrap it much quicker than prior chips. That said, there are a slew of new features in Tegra194, not all of which are easy to upstream. But we're still aiming for at least feature parity with prior chips and get as much of the new hardware supported as we can. Thierry
Mesa support aside- if I start a computationally intensive job on the Jetson TX2 like building the Linux kernel on all cores, it will lock up. My only work around has been to disable the Denver CPU's. I don't think the tegra186 has upstream support to control the fan on the Jetson TX2, could this be a thermal problem? On Fri, Aug 3, 2018, 6:43 AM Thierry Reding <thierry.reding@gmail.com> wrote: > > On Sat, Jul 14, 2018 at 02:22:10PM -0700, Olof Johansson wrote: > > On Thu, Jul 12, 2018 at 05:41:28PM +0200, Thierry Reding wrote: > > > Hi ARM SoC maintainers, > > > > > > The following changes since commit ce397d215ccd07b8ae3f71db689aedb85d56ab40: > > > > > > Linux 4.18-rc1 (2018-06-17 08:04:49 +0900) > > > > > > are available in the Git repository at: > > > > > > git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux.git tags/tegra-for-4.19-arm64-dt > > > > > > for you to fetch changes up to 7780a03495e13cd2bef704bcbf8c727de9f65232: > > > > > > arm64: tegra: Add CPU nodes to Tegra194 device tree (2018-07-02 15:57:39 +0200) > > > > > > Thanks, > > > Thierry > > > > > > ---------------------------------------------------------------- > > > arm64: tegra: Device tree changes for v4.19-rc1 > > > > > > These changes enable the GPIO controllers on Tegra194 SoCs, which in > > > turn allows the SD card detection and ethernet controllers to be enabled > > > as well. The Tegra194 device tree is also extended with the list of CPUs > > > and a PSCI node to inform the kernel about the presence of PSCI capable > > > firmware. > > > > Merged, thanks. Are you planning on supporting 194 better than 186? That never > > really made it to a useful level upstream I think? :( > > I think we have pretty decent upstream support for Tegra186 at this > point. It took a little time because Tegra186 was a little work than > many previous chips because the changes were less incremental than > before. > > Generally our goal is to provide at least feature parity with prior > chips for each new chip. Usually we manage to incrementally add new > features, too. > > Perhaps the reason why it seems that Tegra186 is not well supported > upstream is because a lot of things have moved into BPMP firmware, > which deals with a lot of the nitty gritty details that prior chips > needed a driver for. > > That said, I may be missing something. Do you have any specific features > in mind that aren't working upstream on Tegra186? > > Tegra194 uses the same firmware as Tegra186 (well, a binary compatible > interface to the firmware, I should say), so we've been able to > bootstrap it much quicker than prior chips. That said, there are a slew > of new features in Tegra194, not all of which are easy to upstream. But > we're still aiming for at least feature parity with prior chips and get > as much of the new hardware supported as we can. > > Thierry
On Fri, Aug 03, 2018 at 07:26:04AM -0400, Anthony Eden wrote: > Mesa support aside- if I start a computationally intensive job on the > Jetson TX2 like building the Linux kernel on all cores, it will lock > up. My only work around has been to disable the Denver CPU's. I don't > think the tegra186 has upstream support to control the fan on the > Jetson TX2, could this be a thermal problem? Yes, I suppose this could be a thermal problem. Or it could be something else entirely. We do support CPU frequency scaling on Tegra X2, so what you could do is keep the Denver CPUs enabled, but set the powersave CPU frequency governor. That way it should use all the CPUs but at a lower clock rate, which should also be able to avoid any thermal issues. This could help determine whether or not the problem is thermal or something else. Also adding Mikko on Cc who wrote the Tegra186 driver, maybe he's aware of any issues. Thierry
On 09.08.2018 13:21, Thierry Reding wrote: > On Fri, Aug 03, 2018 at 07:26:04AM -0400, Anthony Eden wrote: >> Mesa support aside- if I start a computationally intensive job on the >> Jetson TX2 like building the Linux kernel on all cores, it will lock >> up. My only work around has been to disable the Denver CPU's. I don't >> think the tegra186 has upstream support to control the fan on the >> Jetson TX2, could this be a thermal problem? > > Yes, I suppose this could be a thermal problem. Or it could be something > else entirely. We do support CPU frequency scaling on Tegra X2, so what > you could do is keep the Denver CPUs enabled, but set the powersave CPU > frequency governor. That way it should use all the CPUs but at a lower > clock rate, which should also be able to avoid any thermal issues. This > could help determine whether or not the problem is thermal or something > else. > > Also adding Mikko on Cc who wrote the Tegra186 driver, maybe he's aware > of any issues. I haven't seen any issues myself, though I haven't stressed the CPU too heavily. We also have a thermal driver for Tegra186, so we could set up thermal throttling with a device tree change. > > Thierry >
On Thu, Aug 09, 2018 at 01:34:37PM +0300, Mikko Perttunen wrote: > On 09.08.2018 13:21, Thierry Reding wrote: > > On Fri, Aug 03, 2018 at 07:26:04AM -0400, Anthony Eden wrote: > > > Mesa support aside- if I start a computationally intensive job on the > > > Jetson TX2 like building the Linux kernel on all cores, it will lock > > > up. My only work around has been to disable the Denver CPU's. I don't > > > think the tegra186 has upstream support to control the fan on the > > > Jetson TX2, could this be a thermal problem? > > > > Yes, I suppose this could be a thermal problem. Or it could be something > > else entirely. We do support CPU frequency scaling on Tegra X2, so what > > you could do is keep the Denver CPUs enabled, but set the powersave CPU > > frequency governor. That way it should use all the CPUs but at a lower > > clock rate, which should also be able to avoid any thermal issues. This > > could help determine whether or not the problem is thermal or something > > else. > > > > Also adding Mikko on Cc who wrote the Tegra186 driver, maybe he's aware > > of any issues. > > I haven't seen any issues myself, though I haven't stressed the CPU too > heavily. We also have a thermal driver for Tegra186, so we could set up > thermal throttling with a device tree change. Do you have an example of how that would work? The DT bindings are a little sparse on the specifics. It seems like something similar to what we did on Tegra124 could be done on Tegra186. Anthony: do you think you could come up with something suitable based on what arch/arm/boot/dts/tegra124{.dtsi,-jetson-tk1.dts} and the device tree bindings for Tegra186 contain in Documentation/devicetree/bindings/thermal/nvidia,tegra186-bpmp-thermal.txt as well as include/dt-bindings/thermal/tegra186-bpmp-thermal.h ? That's provided that reducing the CPU frequency does indeed prevent the lock up that you were seeing. Thierry
Sorry for the late reply. Thank you for the helpful information and guidance. But before I investigate the thermal hypothesis further, I thought I'd send out a kernel panic that I captured today during one of these hangs. At the time I was upgrading packages via pacman (ArchLinux). Does this shed any light on the issue? Best, -Anthony On Thu, Aug 9, 2018 at 10:07 AM Thierry Reding <thierry.reding@gmail.com> wrote: > > On Thu, Aug 09, 2018 at 01:34:37PM +0300, Mikko Perttunen wrote: > > On 09.08.2018 13:21, Thierry Reding wrote: > > > On Fri, Aug 03, 2018 at 07:26:04AM -0400, Anthony Eden wrote: > > > > Mesa support aside- if I start a computationally intensive job on the > > > > Jetson TX2 like building the Linux kernel on all cores, it will lock > > > > up. My only work around has been to disable the Denver CPU's. I don't > > > > think the tegra186 has upstream support to control the fan on the > > > > Jetson TX2, could this be a thermal problem? > > > > > > Yes, I suppose this could be a thermal problem. Or it could be something > > > else entirely. We do support CPU frequency scaling on Tegra X2, so what > > > you could do is keep the Denver CPUs enabled, but set the powersave CPU > > > frequency governor. That way it should use all the CPUs but at a lower > > > clock rate, which should also be able to avoid any thermal issues. This > > > could help determine whether or not the problem is thermal or something > > > else. > > > > > > Also adding Mikko on Cc who wrote the Tegra186 driver, maybe he's aware > > > of any issues. > > > > I haven't seen any issues myself, though I haven't stressed the CPU too > > heavily. We also have a thermal driver for Tegra186, so we could set up > > thermal throttling with a device tree change. > > Do you have an example of how that would work? The DT bindings are a > little sparse on the specifics. It seems like something similar to what > we did on Tegra124 could be done on Tegra186. > > Anthony: do you think you could come up with something suitable based on > what arch/arm/boot/dts/tegra124{.dtsi,-jetson-tk1.dts} and the device > tree bindings for Tegra186 contain in > > Documentation/devicetree/bindings/thermal/nvidia,tegra186-bpmp-thermal.txt > > as well as > > include/dt-bindings/thermal/tegra186-bpmp-thermal.h > > ? That's provided that reducing the CPU frequency does indeed prevent > the lock up that you were seeing. > > Thierry /usr/lib/systemd/systemd: error wh[ 7.411931] Kernel panic - not syncing: Attempted to kill init! exitcode=0x00007f00 [ 7.411931] [ 7.423817] CPU: 0 PID: 1 Comm: systemd Tainted: G S 4.19.0-22-ARCH #1 [ 7.431661] Hardware name: NVIDIA Tegra186 P2771-0000 Development Board (DT) [ 7.438721] Call trace: [ 7.441176] dump_backtrace+0x0/0x180 [ 7.444845] show_stack+0x24/0x30 [ 7.448168] dump_stack+0x9c/0xbc [ 7.451490] panic+0x124/0x274 [ 7.454551] do_exit+0xa80/0xab0 [ 7.457784] do_group_exit+0x3c/0xd0 [ 7.461365] __arm64_sys_exit_group+0x24/0x28 [ 7.465729] el0_svc_common+0x94/0xe8 [ 7.469397] el0_svc_handler+0x38/0x80 [ 7.473152] el0_svc+0x8/0xc [ 7.476039] SMP: stopping secondary CPUs [ 7.479974] Kernel Offset: disabled [ 7.483469] CPU features: 0x0,20002000 [ 7.487222] Memory Limit: none [ 7.490285] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00007f00 [ 7.490285] ]--- ile loading shared libraries: /u[ 7.500730] WARNING: CPU: 0 PID: 1 at kernel/sched/core.c:1163 set_task_cpu+0x1b8/0x1c8 [ 7.511448] Modules linked in: nvme nvme_core broadcom max77620_wdt bcm_phy_lib max77620_thermal ina3221 tegra_drm drm_kms_helper drm drm_panel_orientation_quirks syscopyarea gpio_keys sysfillrect sysimgblt tegra_bpmp_thermal dwmac_dwc_qos_eth i2c_tegra_bpmp fb_sys_fops stmmac_platform stmmac i2c_tegra host1x [ 7.538902] CPU: 0 PID: 1 Comm: systemd Tainted: G S 4.19.0-22-ARCH #1 [ 7.546748] Hardware name: NVIDIA Tegra186 P2771-0000 Development Board (DT) [ 7.553809] pstate: 20000085 (nzCv daIf -PAN -UAO) [ 7.558609] pc : set_task_cpu+0x1b8/0x1c8 [ 7.562627] lr : try_to_wake_up+0x190/0x478 [ 7.566815] sp : ffff000008003d10 [ 7.570134] x29: ffff000008003d10 x28: ffff0000096160c0 [ 7.575456] x27: ffff0000095fc000 x26: 0000000000000100 [ 7.580779] x25: 0000000000000005 x24: ffff00000961a490 [ 7.586102] x23: ffff0000096089c0 x22: 0000000000000000 [ 7.593268] x21: 0000000000000004 x20: 0000000000000005 [ 7.600426] x19: ffff8001ed1f5e80 x18: 0000000000000000 [ 7.607584] x17: 0000000000000000 x16: 0000000000000000 [ 7.614740] x15: 0000000000000000 x14: 0000000000000000 [ 7.621866] x13: ffff000008ca2658 x12: 00000000ffffffff [ 7.629006] x11: 000000000000009c x10: 0000000000000001 [ 7.636135] x9 : 0000000000000000 x8 : ffff8001f67412a8 [ 7.643241] x7 : 0040000000000000 x6 : 0000000000000036 [ 7.650358] x5 : 00008001ed140000 x4 : ffff00000961a490 [ 7.657457] x3 : 00008001ed1b8000 x2 : 0000000000000005 [ 7.664563] x1 : ffff000009619700 x0 : 0000000000000000 [ 7.671641] Call trace: [ 7.675753] set_task_cpu+0x1b8/0x1c8 [ 7.681081] try_to_wake_up+0x190/0x478 [ 7.686593] wake_up_process+0x28/0x38 [ 7.691993] process_timeout+0x20/0x30 [ 7.697355] call_timer_fn+0x34/0x170 [ 7.702636] expire_timers+0xc0/0x148 [ 7.707908] run_timer_softirq+0xbc/0x1d8 [ 7.713515] __do_softirq+0x120/0x300 [ 7.718781] irq_exit+0xc0/0xd0 [ 7.723505] __handle_domain_irq+0x70/0xc0 [ 7.729138] gic_handle_irq+0x58/0xa8 [ 7.734332] el1_irq+0xb0/0x140 [ 7.739006] panic+0x224/0x274 [ 7.743561] do_exit+0xa80/0xab0 [ 7.748299] do_group_exit+0x3c/0xd0 [ 7.753361] __arm64_sys_exit_group+0x24/0x28 [ 7.759217] el0_svc_common+0x94/0xe8 [ 7.764357] el0_svc_handler+0x38/0x80 [ 7.769562] el0_svc+0x8/0xc [ 7.773915] ---[ end trace 22e2a84658d004da ]--- sr/lib/libcryptsetup.so.12: file too short
Hi ARM SoC maintainers, The following changes since commit ce397d215ccd07b8ae3f71db689aedb85d56ab40: Linux 4.18-rc1 (2018-06-17 08:04:49 +0900) are available in the Git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux.git tags/tegra-for-4.19-arm64-dt for you to fetch changes up to 7780a03495e13cd2bef704bcbf8c727de9f65232: arm64: tegra: Add CPU nodes to Tegra194 device tree (2018-07-02 15:57:39 +0200) Thanks, Thierry ---------------------------------------------------------------- arm64: tegra: Device tree changes for v4.19-rc1 These changes enable the GPIO controllers on Tegra194 SoCs, which in turn allows the SD card detection and ethernet controllers to be enabled as well. The Tegra194 device tree is also extended with the list of CPUs and a PSCI node to inform the kernel about the presence of PSCI capable firmware. ---------------------------------------------------------------- Mikko Perttunen (4): arm64: tegra: Add GPIO controller on Tegra194 arm64: tegra: Enable card detect for SD card on P2888 arm64: tegra: Add ethernet controller on Tegra194 arm64: tegra: Add CPU nodes to Tegra194 device tree arch/arm64/boot/dts/nvidia/tegra194-p2888.dtsi | 22 ++++- arch/arm64/boot/dts/nvidia/tegra194.dtsi | 106 +++++++++++++++++++++++++ 2 files changed, 126 insertions(+), 2 deletions(-)