On 26.09.2018 21:32, Guenter Roeck wrote: > On Wed, Sep 26, 2018 at 06:36:08PM +0200, Helge Deller wrote: >> On 26.09.2018 15:29, Guenter Roeck wrote: >>> On 09/26/2018 05:09 AM, John David Anglin wrote: >>>> On 2018-09-25 11:21 PM, Guenter Roeck wrote: >>>>> This patch causes my parisc qemu tests to fail. >>>>> Unfortunately I don't have any useful log output; the failure >>>>> is silent. Reverting the patch fixes the problem. >>>> Can you be more specific on how to run these tests? >>> >>> Sorry. Please see >>> >>> https://github.com/groeck/linux-build-test/tree/master/rootfs/parisc >>> >>> My tests enable a number of device and debug options on top of defconfig. >>> Those are not necessary, though. The problem can be reproduced with defconfig. >>> >>> With the initrd available from there, and with an image built using 'defconfig', >>> run >>> >>> qemu-system-hppa \ >>> -kernel vmlinux -no-reboot \ >>> -initrd rootfs.cpio.gz \ >>> -append 'rdinit=/sbin/init console=ttyS0,115200' \ >>> -nographic -monitor null >>> >>> I tested with qemu 2.12 and 3.0. Using "arch/parisc/boot/bzImage" as kernel >>> image does not make a difference. >>> >>> Note: The initrd auto-reboots. To avoid that, add "noreboot" as additional >>> command line option. >> >> You didn't mention which kernel version you were using. > > Sorry. The patch is only available in -next, so I somehow thought it was > obvious that I referred to -next. Mainline as of v4.19-rc5-143-gc307aaf3eb47 > is fine. However, I do observe the problem in mainline after applying this > patch on top of c307aaf3eb47. I also see the problem with your for-next > branch. I tried my -for-next on top of c307aaf3eb47 too -> no issues. >> I did some testing with your rootfs & initrd, based on upstream git kernel >> with the latest patches from my for-next tree: >> https://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux.git/log/?h=for-next >> >> I did tested with my own 32-bit config and the 32bit-default-config. >> >> Overall I couldn't reproduce any issues. > > Interesting. > >> Can you try to come up with some more info for what I should look at? >> > > All I can say is that it doesn't boot for me if this patch is applied. > >> By the way, this patch is now running on all physical debian-buildd servers (64bit kernel) >> without any issues so far. >> > > I guess qemu is at fault, but I have no idea why. I also tried qemu > mainline, but it does not make a difference for me. Do you by any > chance know the command line necessary to make earlycon work ? > Maybe that would help. You could try append "console=ttyB" See arch/parisc/kernel/pdc_cons.c I haven't used in since a long time myself though.... > Looking into debug output from qemu, I see that start_firmware calls some > external function. In the working image, execution returns to start_firmware. > In the non-working image, it never returns but stops execution. > Here is the key difference: > > Working: > > ---------------- > IN: > 0x0000000000175744: extrw,u r8,9,10,r1 > 0x0000000000175748: depw r0,31,12,r25 > 0x000000000017574c: copy r0,r16 > 0x0000000000175750: ldw,s r1(r25),r25 > 0x0000000000175754: bb,>=,n r25,1f,0x1757cc > > ---------------- > IN: > 0x0000000000175758: depw r0,31,4,r25 > 0x000000000017575c: copy r25,r9 > 0x0000000000175760: depw,z r9,23,24,r25 > 0x0000000000175764: extrw,u r8,19,10,r1 > 0x0000000000175768: depw r0,31,12,r25 > 0x000000000017576c: shladd r1,2,r25,r25 > 0x0000000000175770: ldw r0(r25),r16 <<-- not executed in broken case > 0x0000000000175774: bb,>=,n r16,16,0x1757cc <<-- not executed in broken case Yes, it's not executed then, because it was dropped with the patch. > ---------------- > IN: > 0x0000000000175778: ldi 100,r9 > 0x000000000017577c: or r9,r16,r1 > 0x0000000000175780: and,<> r9,r16,r0 > 0x0000000000175784: stw r1,0(r25) > 0x0000000000175788: depw,z r24,30,15,r17 > 0x000000000017578c: depw r16,8,7,r17 > 0x0000000000175790: extrw,u,= r16,24,1,r0 > 0x0000000000175794: depwi 1,12,1,r17 > 0x0000000000175798: extrw,u,= r16,20,1,r0 > 0x000000000017579c: depwi 7,11,3,r17 > 0x00000000001757a0: extrw,u,= r16,28,1,r0 > 0x00000000001757a4: depwi 0,11,2,r17 > 0x00000000001757a8: depwi 0,31,12,r16 > 0x00000000001757ac: extrw,u r16,24,25,r16 > 0x00000000001757b0: mfsp sr1,r9 > 0x00000000001757b4: mtsp r24,sr1 > 0x00000000001757b8: idtlba r16,(sr1,r8) > 0x00000000001757bc: idtlbp r17,(sr1,r8) > 0x00000000001757c0: mtsp r9,sr1 > 0x00000000001757c4: rfi,r > > ---------------- > IN: start_parisc > 0x0000000010107694: copy r3,r1 > 0x0000000010107698: stw rp,-14(sp) > > ... > > Broken: > > ---------------- > IN: > 0x0000000000175744: extrw,u r8,9,10,r1 > 0x0000000000175748: depw r0,31,12,r25 > 0x000000000017574c: copy r0,r16 > 0x0000000000175750: ldw,s r1(r25),r25 > 0x0000000000175754: bb,>=,n r25,1f,0x1757c4 <<-- shightly different jump target Code above is from arch/parisc/kernel/entry.S:427 (L2_ptep macro), so jump target is different. > > ---------------- > IN: > 0x0000000000175758: depw r0,31,4,r25 > 0x000000000017575c: copy r25,r9 > 0x0000000000175760: depw,z r9,23,24,r25 > 0x0000000000175764: extrw,u r8,19,10,r1 > 0x0000000000175768: depw r0,31,12,r25 > 0x000000000017576c: shladd r1,2,r25,r25 > <<-- two instructions missing > 0x0000000000175770: ldi 100,r9 > 0x0000000000175774: or r9,r16,r1 > 0x0000000000175778: and,<> r9,r16,r0 > 0x000000000017577c: stw r1,0(r25) > 0x0000000000175780: depw,z r24,30,15,r17 > 0x0000000000175784: depw r16,8,7,r17 > 0x0000000000175788: extrw,u,= r16,24,1,r0 > 0x000000000017578c: depwi 1,12,1,r17 > 0x0000000000175790: extrw,u,= r16,20,1,r0 > 0x0000000000175794: depwi 7,11,3,r17 > 0x0000000000175798: extrw,u,= r16,28,1,r0 > 0x000000000017579c: depwi 0,11,2,r17 > 0x00000000001757a0: depwi 0,31,12,r16 > 0x00000000001757a4: extrw,u r16,24,25,r16 > 0x00000000001757a8: mfsp sr1,r9 > 0x00000000001757ac: mtsp r24,sr1 > 0x00000000001757b0: idtlba r16,(sr1,r8) > 0x00000000001757b4: idtlbp r17,(sr1,r8) > 0x00000000001757b8: mtsp r9,sr1 > 0x00000000001757bc: rfi,r <<-- should return to start_parisc > > ---------------- > IN: > 0x00000000001738e0: b,l 0x1754bc,r0 > 0x00000000001738e4: ldi 7,r8 > ... > > ---------------- > IN: > 0x00000000001755e8: mtctl r0,pcsq > 0x00000000001755ec: mtctl r0,pcsq > 0x00000000001755f0: mtctl r1,ipsw > 0x00000000001755f4: ldil L%10175000,r1 > 0x00000000001755f8: ldo 610(r1),r1 > 0x00000000001755fc: mtctl r1,pcoq > 0x0000000000175600: ldo 4(r1),r1 > 0x0000000000175604: mtctl r1,pcoq > 0x0000000000175608: rfi,r <<-- hangs here > > > The differences seem to be in line with the patch. > > No idea how to proceed. Please let me know if there is anything else I can do. Sorry, I don't see any obvious wrong and am currently out of ideas... Helge
