Em Tue, 29 Sep 2015 16:32:58 +0200 Sylwester Nawrocki <s.nawrocki@xxxxxxxxxxx> escreveu: > Adding linux-samsung-soc mailing list at Cc. Never mind. My mistake: I forgot to reinstall the modules (mwifiex and mfc drivers were compiled as such). It is working. I guess the only remaining thing is to add the dependency of EXYNOS_IOMMU to the MFC driver. Regards, Mauro > > On 29/09/15 16:01, Mauro Carvalho Chehab wrote: > > Em Tue, 29 Sep 2015 15:37:51 +0200 > > Sylwester Nawrocki <s.nawrocki@xxxxxxxxxxx> escreveu: > > > >> Hi Mauro, > >> > >> On 29/09/15 14:43, Mauro Carvalho Chehab wrote: > >>> Hi Sylwester, > >>> > >>> I'm c/c Javier. Javier is working with us at the Open Souce Group and he > >>> has been working on making sure that Chromebooks are supported with vanilla > >>> upstream Kernels. > >>> > >>> That's said, while testing the latest upstream Kernel (4.3-rc3) on a > >>> Chromebook snow, I noticed this error message: > >>> > >>> [ 211.116975] s5p_mfc_alloc_memdevs:1045: Failed to declare coherent memory for > >>> MFC device > >>> [ 211.117039] s5p-mfc: probe of 11000000.codec failed with error -12 > >>> > >>> Do you have any idea about how to fix? Btw, do I need any firmware to test > >>> the MFC driver there? > >> > >> You can get v4.1 based kernel with MFC working (e.g. on Odroid XU3) here: > >> > >> https://USER@xxxxxxxxxxxxxxxx/gerrit/p/platform/kernel/linux-exynos.git > >> ssh://USER@xxxxxxxxxxxxxxxx:29418/platform/kernel/linux-exynos.git > >> branch: tizen > >> > >> I'm adding at Cc Marek who knows more details. AFAIU you may need patches > >> like these on top of current mainline kernel: > >> > >> https://review.tizen.org/gerrit/gitweb?p=platform/kernel/linux-exynos.git;a=commitdiff;h=dcbf130658cdd1731b8a219ecedd867af0b2670e > >> > >> https://review.tizen.org/gerrit/gitweb?p=platform/kernel/linux-exynos.git;a=commitdiff;h=b82fd3408fd58476da79996d4a3bcbbb72174961 > >> > >> And Exynos IOMMU needs to be enabled in kernel config. > > > > If this is a requirement, we should enforce it at s5p-mfc Kconfig entry, > > like: > > > > depends on EXYNOS_IOMMU if !COMPILE_TEST > > In general s5p-mfc doesn't depend on IOMMU, some SoC revisions don't > even have SYSMMU. Thus we can't add such constraint. > > > Just enabled Exynos IOMMU here to see what would happen: > > > > [ 19.461839] ------------[ cut here ]------------ > > [ 19.464264] kernel BUG at net/core/dev.c:6513! > > [ 19.466648] Internal error: Oops - BUG: 0 [#2] PREEMPT SMP ARM > > [ 19.469039] Modules linked in: mwifiex_sdio mwifiex exynos_gsc(+) v4l2_mem2mem s5p_mfc(+) videobuf2_dma_contig videobuf2_memops videobuf2_core v4l2_common videodev media > > [ 19.474158] CPU: 1 PID: 1390 Comm: kworker/1:2 Tainted: G D 4.3.0-rc3-00492-g397ead8befa9 #10 > > [ 19.476753] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree) > > [ 19.479326] Workqueue: events request_firmware_work_func > > [ 19.481887] task: edb06d00 ti: ee18a000 task.ti: ee18a000 > > [ 19.484445] PC is at register_netdevice+0x23c/0x41c > > [ 19.487013] LR is at register_netdevice+0x28/0x41c > > [ 19.489572] pc : [<c047a184>] lr : [<c0479f70>] psr: 60010013 > > [ 19.489572] sp : ee18be00 ip : 0000000e fp : ed4629a0 > > [ 19.494692] r10: bf102940 r9 : 00000000 r8 : ed462000 > > [ 19.497218] r7 : 00000002 r6 : ed462000 r5 : edbee000 r4 : ed738000 > > [ 19.499741] r3 : 00000000 r2 : 00000146 r1 : 000003e8 r0 : 00000001 > > [ 19.502253] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none > > [ 19.504768] Control: 10c5387d Table: 6d5bc06a DAC: 00000051 > > [ 19.507247] Process kworker/1:2 (pid: 1390, stack limit = 0xee18a210) > > [ 19.509717] Stack: (0xee18be00 to 0xee18c000) > > [ 19.512150] be00: 0000ffff 00000001 00000000 00000000 ed4629a0 ed738000 edbee000 ed462000 > > [ 19.514615] be20: 00000002 ed7392fc ed739000 bf0fab4c 00000004 00000001 edbef000 ffffffff > > [ 19.517072] be40: c086697c edbee000 bf10b400 edbef000 bf12d784 00000000 00000000 ed9a4800 > > [ 19.519518] be60: ee7b58c0 bf0d95b0 00000000 00000000 00000000 00000000 f1694000 0006fd7c > > [ 19.521959] be80: 00000000 ec118e00 ec118dc0 c02f711c f1694000 0006fd7c ed43a390 ec118d40 > > [ 19.524396] bea0: 00000003 ee18bee4 edbbf208 c0834674 ec118dc0 c08ba8bc 00000002 ec118d80 > > [ 19.526827] bec0: ed9a4800 ec118d80 > > > > Unfortunately, the serial console didn't got everything. It _seems_ that the > > bug happens during a mwifiex kthread, when it calls mwifiex_add_virtual_intf, > > but even at the device's screen the message is truncated. > > Unfortunately this doesn't ring any bells here, we don't work with Chromebooks > in general and I'm not sure how to address this issue. Maybe someone subscribed > to the mailing list can help. > > > I didn't apply those two patches you pointed from Tizen.org. Not sure if > > this is related or not. > > It's possible the issue is caused by some other device (driver) that also > has a sysmmu assigned to it, like display controller, etc. I'm not sure. > > >> Required firmware is already available the linux-firmware git repository, > >> the driver will select proper version if the firmware files (s5p-mfc*.fw) > >> are copied to /lib/firmware. > > > > Ah, OK! > >> > >>> Thanks! > >>> Mauro > -- To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
 |