| Message ID | 20191018160735.15658-2-m.tretter@pengutronix.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: zynqmp: add firmware device tree node | expand |
On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: > From: Rajan Vaja <rajan.vaja@xilinx.com> > > Add firmware DT node in ZynqMP device tree. This node > uses bindings as per new firmware interface driver. > > Signed-off-by: Rajan Vaja <rajanv@xilinx.com> > Signed-off-by: Michal Simek <michal.simek@xilinx.com> > Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 emulation crashes (see below). Any idea what it might take to get qemu back to working ? Thanks, Guenter --- [ 30.719268] ------------[ cut here ]------------ [ 30.719403] kernel BUG at arch/arm64/kernel/traps.c:406! [ 30.719971] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP [ 30.720358] Modules linked in: [ 30.720741] CPU: 3 PID: 1 Comm: swapper/0 Not tainted 5.4.0-13331-g9455d25f4e3b #1 [ 30.720852] Hardware name: ZynqMP ZCU102 Rev1.0 (DT) [ 30.721157] pstate: 00000005 (nzcv daif -PAN -UAO) [ 30.721261] pc : do_undefinstr+0x2f4/0x318 [ 30.721336] lr : do_undefinstr+0x1fc/0x318 [ 30.721410] sp : ffff80001003b930 [ 30.721486] x29: ffff80001003b930 x28: ffff00007d178040 [ 30.721597] x27: 0000000000000000 x26: ffff8000117d0514 [ 30.721683] x25: ffff8000118c90d0 x24: 0000000000000000 [ 30.721769] x23: 0000000040000005 x22: 00000000d4000003 [ 30.721854] x21: ffff800011e1c850 x20: ffff80001003b990 [ 30.721940] x19: ffff800011e0fa08 x18: 0000000000000001 [ 30.722025] x17: ffff800010c0f1d8 x16: ffff800010c11fb8 [ 30.722111] x15: ffffffffffffffff x14: ffffffffffffffff [ 30.722196] x13: 0000000000000018 x12: 0101010101010101 [ 30.722281] x11: 0000000000000000 x10: 00000000628e21fa [ 30.722384] x9 : ffff00007d178858 x8 : ffff00007d178880 [ 30.722471] x7 : ffff80001003b8b0 x6 : 0000000000000001 [ 30.722560] x5 : 0000000000000001 x4 : 0000000000000001 [ 30.722646] x3 : 0000000000000000 x2 : 00000000000174b1 [ 30.722730] x1 : ffff00007d178040 x0 : 0000000040000005 [ 30.722913] Call trace: [ 30.722993] do_undefinstr+0x2f4/0x318 [ 30.723070] el1_sync_handler+0xb0/0x108 [ 30.723138] el1_sync+0x7c/0x100 [ 30.723201] __arm_smccc_smc+0x0/0x2c [ 30.723272] zynqmp_pm_get_api_version.part.1+0x40/0x68 [ 30.723352] zynqmp_firmware_probe+0xbc/0x298
Hi, +Edgar On 08. 12. 19 23:38, Guenter Roeck wrote: > On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >> From: Rajan Vaja <rajan.vaja@xilinx.com> >> >> Add firmware DT node in ZynqMP device tree. This node >> uses bindings as per new firmware interface driver. >> >> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> > > With this patch applied in the mainline kernel, the qemu xlnx-zcu102 > emulation crashes (see below). Any idea what it might take to get > qemu back to working ? Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB concept is working with mainline qemu. But crash is too hard. It should be no response from PMU and then this panic. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/firmware/xilinx/zynqmp.c?h=v5.5-rc1#n728 Edgar: can you please comment this? Thanks, Michal
On 12/8/19 10:42 PM, Michal Simek wrote: > Hi, +Edgar > > > On 08. 12. 19 23:38, Guenter Roeck wrote: >> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>> >>> Add firmware DT node in ZynqMP device tree. This node >>> uses bindings as per new firmware interface driver. >>> >>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >> >> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >> emulation crashes (see below). Any idea what it might take to get >> qemu back to working ? > > Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB > concept is working with mainline qemu. But crash is too hard. It should > be no response from PMU and then this panic. > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/firmware/xilinx/zynqmp.c?h=v5.5-rc1#n728 > Isn't that a bit harsh too ? Normally one would print an error message and abort driver instantiation. It sounds like you are saying that qemu's xlnx-zcu102 emulation is no longer supported and expected to crash the kernel. Is this a correct assumption ? If so, I'll drop it from my list of tests. Thanks, Guenter
On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: > On 12/8/19 10:42 PM, Michal Simek wrote: > > Hi, +Edgar > > > > > > On 08. 12. 19 23:38, Guenter Roeck wrote: > > > On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: > > > > From: Rajan Vaja <rajan.vaja@xilinx.com> > > > > > > > > Add firmware DT node in ZynqMP device tree. This node > > > > uses bindings as per new firmware interface driver. > > > > > > > > Signed-off-by: Rajan Vaja <rajanv@xilinx.com> > > > > Signed-off-by: Michal Simek <michal.simek@xilinx.com> > > > > Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> > > > > > > With this patch applied in the mainline kernel, the qemu xlnx-zcu102 > > > emulation crashes (see below). Any idea what it might take to get > > > qemu back to working ? > > > > Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB > > concept is working with mainline qemu. But crash is too hard. It should Yes, QEMU doesn't support the Cortex-A53s along with the PMU MicroBlaze. My workaround when using upstream QEMU is a modified DT without the PMU firmware and with fixed-clock nodes. > > be no response from PMU and then this panic. > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/firmware/xilinx/zynqmp.c?h=v5.5-rc1#n728 > > > > Isn't that a bit harsh too ? Normally one would print an error message > and abort driver instantiation. I agree, it would be nice if ATF & kernel drivers would somehow handle this more gracefully. Cheers, Edgar > > It sounds like you are saying that qemu's xlnx-zcu102 emulation is > no longer supported and expected to crash the kernel. Is this a > correct assumption ? If so, I'll drop it from my list of tests. > > Thanks, > Guenter
On 09. 12. 19 8:48, Edgar E. Iglesias wrote: > On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: >> On 12/8/19 10:42 PM, Michal Simek wrote: >>> Hi, +Edgar >>> >>> >>> On 08. 12. 19 23:38, Guenter Roeck wrote: >>>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>>> >>>>> Add firmware DT node in ZynqMP device tree. This node >>>>> uses bindings as per new firmware interface driver. >>>>> >>>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>>> >>>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>>> emulation crashes (see below). Any idea what it might take to get >>>> qemu back to working ? >>> >>> Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB >>> concept is working with mainline qemu. But crash is too hard. It should > > Yes, QEMU doesn't support the Cortex-A53s along with the PMU MicroBlaze. > > My workaround when using upstream QEMU is a modified DT without the PMU firmware > and with fixed-clock nodes. IIRC you said that there is still discussion how to upstream this. Fixed clock should work for u-boot too. But SPL reads that registers directly. Are you implementing them with any default values? > > >>> be no response from PMU and then this panic. >>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/firmware/xilinx/zynqmp.c?h=v5.5-rc1#n728 >>> >> >> Isn't that a bit harsh too ? Normally one would print an error message >> and abort driver instantiation. > > I agree, it would be nice if ATF & kernel drivers would somehow handle > this more gracefully. Rajan: can you please take a look at it? Thanks, Michal
On 09. 12. 19 8:19, Guenter Roeck wrote: > On 12/8/19 10:42 PM, Michal Simek wrote: >> Hi, +Edgar >> >> >> On 08. 12. 19 23:38, Guenter Roeck wrote: >>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>> >>>> Add firmware DT node in ZynqMP device tree. This node >>>> uses bindings as per new firmware interface driver. >>>> >>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>> >>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>> emulation crashes (see below). Any idea what it might take to get >>> qemu back to working ? >> >> Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB >> concept is working with mainline qemu. But crash is too hard. It should >> be no response from PMU and then this panic. >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/firmware/xilinx/zynqmp.c?h=v5.5-rc1#n728 >> >> > > Isn't that a bit harsh too ? Normally one would print an error message > and abort driver instantiation. Using different DT as Edgar mentioned should work but the question is how to properly wire it up. Edgar: What about "simply" generate that DT by qemu model based on what implementation is available? > > It sounds like you are saying that qemu's xlnx-zcu102 emulation is > no longer supported and expected to crash the kernel. Is this a > correct assumption ? If so, I'll drop it from my list of tests. Let's discuss this first. I definitely appreciate that you run these tests. Thanks, Michal
On 12/8/19 11:48 PM, Edgar E. Iglesias wrote: > On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: >> On 12/8/19 10:42 PM, Michal Simek wrote: >>> Hi, +Edgar >>> >>> >>> On 08. 12. 19 23:38, Guenter Roeck wrote: >>>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>>> >>>>> Add firmware DT node in ZynqMP device tree. This node >>>>> uses bindings as per new firmware interface driver. >>>>> >>>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>>> >>>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>>> emulation crashes (see below). Any idea what it might take to get >>>> qemu back to working ? >>> >>> Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB >>> concept is working with mainline qemu. But crash is too hard. It should > > Yes, QEMU doesn't support the Cortex-A53s along with the PMU MicroBlaze. > > My workaround when using upstream QEMU is a modified DT without the PMU firmware > and with fixed-clock nodes. > I can't do that for my boot tests. Normally I would just disable ZYNQMP_FIRMWARE, but that is hard enabled with ARCH_ZYNQMP. I'll have to drop those tests, unfortunately, if the firmware driver is considered mandatory. Guenter
On 09. 12. 19 15:32, Guenter Roeck wrote: > On 12/8/19 11:48 PM, Edgar E. Iglesias wrote: >> On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: >>> On 12/8/19 10:42 PM, Michal Simek wrote: >>>> Hi, +Edgar >>>> >>>> >>>> On 08. 12. 19 23:38, Guenter Roeck wrote: >>>>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>>>> >>>>>> Add firmware DT node in ZynqMP device tree. This node >>>>>> uses bindings as per new firmware interface driver. >>>>>> >>>>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>>>> >>>>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>>>> emulation crashes (see below). Any idea what it might take to get >>>>> qemu back to working ? >>>> >>>> Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB >>>> concept is working with mainline qemu. But crash is too hard. It should >> >> Yes, QEMU doesn't support the Cortex-A53s along with the PMU MicroBlaze. >> >> My workaround when using upstream QEMU is a modified DT without the >> PMU firmware >> and with fixed-clock nodes. >> > > I can't do that for my boot tests. Normally I would just disable > ZYNQMP_FIRMWARE, > but that is hard enabled with ARCH_ZYNQMP. I'll have to drop those tests, > unfortunately, if the firmware driver is considered mandatory. We can make it optional. Rajan: please send a patch for it. M
On 12/9/19 7:02 AM, Michal Simek wrote: > On 09. 12. 19 15:32, Guenter Roeck wrote: >> On 12/8/19 11:48 PM, Edgar E. Iglesias wrote: >>> On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: >>>> On 12/8/19 10:42 PM, Michal Simek wrote: >>>>> Hi, +Edgar >>>>> >>>>> >>>>> On 08. 12. 19 23:38, Guenter Roeck wrote: >>>>>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>>>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>>>>> >>>>>>> Add firmware DT node in ZynqMP device tree. This node >>>>>>> uses bindings as per new firmware interface driver. >>>>>>> >>>>>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>>>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>>>>> >>>>>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>>>>> emulation crashes (see below). Any idea what it might take to get >>>>>> qemu back to working ? >>>>> >>>>> Driver talks through ATF to PMU unit(microblaze). I don't think A53+MB >>>>> concept is working with mainline qemu. But crash is too hard. It should >>> >>> Yes, QEMU doesn't support the Cortex-A53s along with the PMU MicroBlaze. >>> >>> My workaround when using upstream QEMU is a modified DT without the >>> PMU firmware >>> and with fixed-clock nodes. >>> >> >> I can't do that for my boot tests. Normally I would just disable >> ZYNQMP_FIRMWARE, >> but that is hard enabled with ARCH_ZYNQMP. I'll have to drop those tests, >> unfortunately, if the firmware driver is considered mandatory. > > We can make it optional. > Rajan: please send a patch for it. > I'll disable the related boot tests for now. If/when this is fixed, let me know, and I'll re-enable it. Guenter
On 15. 12. 19 6:28, Guenter Roeck wrote: > On 12/9/19 7:02 AM, Michal Simek wrote: >> On 09. 12. 19 15:32, Guenter Roeck wrote: >>> On 12/8/19 11:48 PM, Edgar E. Iglesias wrote: >>>> On Sun, Dec 08, 2019 at 11:19:33PM -0800, Guenter Roeck wrote: >>>>> On 12/8/19 10:42 PM, Michal Simek wrote: >>>>>> Hi, +Edgar >>>>>> >>>>>> >>>>>> On 08. 12. 19 23:38, Guenter Roeck wrote: >>>>>>> On Fri, Oct 18, 2019 at 06:07:31PM +0200, Michael Tretter wrote: >>>>>>>> From: Rajan Vaja <rajan.vaja@xilinx.com> >>>>>>>> >>>>>>>> Add firmware DT node in ZynqMP device tree. This node >>>>>>>> uses bindings as per new firmware interface driver. >>>>>>>> >>>>>>>> Signed-off-by: Rajan Vaja <rajanv@xilinx.com> >>>>>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >>>>>>>> Signed-off-by: Michael Tretter <m.tretter@pengutronix.de> >>>>>>> >>>>>>> With this patch applied in the mainline kernel, the qemu xlnx-zcu102 >>>>>>> emulation crashes (see below). Any idea what it might take to get >>>>>>> qemu back to working ? >>>>>> >>>>>> Driver talks through ATF to PMU unit(microblaze). I don't think >>>>>> A53+MB >>>>>> concept is working with mainline qemu. But crash is too hard. It >>>>>> should >>>> >>>> Yes, QEMU doesn't support the Cortex-A53s along with the PMU >>>> MicroBlaze. >>>> >>>> My workaround when using upstream QEMU is a modified DT without the >>>> PMU firmware >>>> and with fixed-clock nodes. >>>> >>> >>> I can't do that for my boot tests. Normally I would just disable >>> ZYNQMP_FIRMWARE, >>> but that is hard enabled with ARCH_ZYNQMP. I'll have to drop those >>> tests, >>> unfortunately, if the firmware driver is considered mandatory. >> >> We can make it optional. >> Rajan: please send a patch for it. >> > > I'll disable the related boot tests for now. If/when this is fixed, let > me know, > and I'll re-enable it. ok. Sure. Thanks, Michal
diff --git a/arch/arm64/boot/dts/xilinx/zynqmp.dtsi b/arch/arm64/boot/dts/xilinx/zynqmp.dtsi index 9aa67340a4d8..9115eaebbf70 100644 --- a/arch/arm64/boot/dts/xilinx/zynqmp.dtsi +++ b/arch/arm64/boot/dts/xilinx/zynqmp.dtsi @@ -115,6 +115,13 @@ method = "smc"; }; + firmware { + zynqmp_firmware: zynqmp-firmware { + compatible = "xlnx,zynqmp-firmware"; + method = "smc"; + }; + }; + timer { compatible = "arm,armv8-timer"; interrupt-parent = <&gic>;