Image location: http://autobuilder.yoctoproject.org/pub/nightly/20130313-3 Branch: poky/master Git rev: 4da406cd50cf02a63ed76810a670aba1eb561c79 test step: 1. bootup qemuarm with "core-image-sato-qemuarm-20130313194304.rootfs.ext3" 2. open terminal from qemuarm 3. execute "dmesg | grep -i error", there should no error list. But it contains two errors like below: mmci-pl18x: probe of fpga:05 failed with error -38 mmci-pl18x: probe of fpga:0b failed with error -38 4. when use "core-image-sato-sdk-qemuarm-20130313194304.rootfs.ext3", same errors: [ 4.430820] mmci-pl18x: probe of fpga:05 failed with error -38 [ 4.443642] mmci-pl18x: probe of fpga:0b failed with error -38 comments: do the same test on qemuppc, no this issue.
This was reported by the AB and fixed with: commit d5267afa634a4da33f5d5a88c678387a9b7ec1b1 Author: Richard Purdie <richard.purdie@linuxfoundation.org> Date: Sat Mar 2 22:31:19 2013 +0000
(In reply to comment #1) > This was reported by the AB and fixed with: > > commit d5267afa634a4da33f5d5a88c678387a9b7ec1b1 > Author: Richard Purdie <richard.purdie@linuxfoundation.org> > Date: Sat Mar 2 22:31:19 2013 +0000 Hi， I saw this fix only modify the test code, to skip mmci-pl18x dmesg. Does this mean we do not care about the mmci-pl18x error info anymore? Since only arm qemu has this issue, ppc no this error. This modification will skip all qemus mmci-pl18x test. Does it mean we do not need test it on other qemus, such as ppc? Thanks.
No, this is really a workaround until a proper fix can be made.
(In reply to comment #3) > No, this is really a workaround until a proper fix can be made. Got it. Then should we reopen this bug to continue tracking this issue until it be really fixed? Thanks.
Issue found in poky master:cb2836fee444bf5f7bd74d0612a4cd8f1b5d2e89. Reopening.
Just want to confirm this is the same issue or different? Did the same exception (grepping for the pl18x) get moved from the qemuimage-tests to the new test framework? Is this something we can really fix in the kernel? (Bruce & Darren Cc'ed)
If I remember correctly, we didn't feel it was appropriate to fix in the kernel as it was really an issue with qemu's implementation of the given hardware. Bruce, was that about right? I believe the proper fix would be to qemu. However, I don't think it will be considered high enough priority to time for and we should probably just ensure the new tests are still ignoring it.
This was discovered in a manual test of the qemu image. The AB tests still ignore it; see the dmesg grep below: 'dmesg | grep -v mmci-pl18x | grep -v "error changing net interface name" | grep -i error' Even though it's ignored in automated tests, the issue is still present.
(In reply to comment #7) > If I remember correctly, we didn't feel it was appropriate to fix in the > kernel as it was really an issue with qemu's implementation of the given > hardware. Bruce, was that about right? > > I believe the proper fix would be to qemu. However, I don't think it will be > considered high enough priority to time for and we should probably just > ensure the new tests are still ignoring it. Correct. As with the interrupt mappings, we want qemu to represent the h/w. So to silence this, we want a qemu fix, instead of hacking around it in the kernel. This is still a low priority, but I suppose we can keep it around :)
Can you re-test this problem.
This issue still remains in 1.7M2 rc1 build http://autobuilder.yoctoproject.org/pub/releases/yocto-1.7_M2.rc1/ sh-4.3# dmesg | grep -i error [ 6.035639] mmci-pl18x: probe of fpga:05 failed with error -22 [ 6.039104] mmci-pl18x: probe of fpga:0b failed with error -22
At this point we're not going to enhance qemu to address this, we're also not going to hack the kernel to disable it, so marking as WONTFIX. The automated version of the QA test already ignore this 'error' message and the manual QA versions of the test should be dropped.