Am 19.11.2015 um 16:16 schrieb Dr. David Alan Gilbert:
> * Peter Maydell (peter.mayd...@linaro.org) wrote:
>> On 19 November 2015 at 14:44, Peter Maydell <peter.mayd...@linaro.org> wrote:
>>> On 19 November 2015 at 13:21, Peter Maydell <peter.mayd...@linaro.org>
>>> wrote:
>>>> On 19 November 2015 at 13:12, Peter Maydell <peter.mayd...@linaro.org>
>>>> wrote:
>>>>> Hi. Unfortunately this failed in 'make check' (x86-64 Linux, debug build):
>>>>>
>>>>> GTESTER check-qtest-i386
>>>>> blkdebug: Suspended request 'A'
>>>>> blkdebug: Resuming request 'A'
>>>>> **
>>>>> ERROR:/home/petmay01/linaro/qemu-for-merges/tests/ide-test.c:656:ide_wait_clear:
>>>>> code should not be reached
>>>>> GTester: last random seed: R02Sf1d98a814dee34e3985479976b17883c
>>>>> main-loop: WARNING: I/O thread spun for 1000 iterations
>>>>> main-loop: WARNING: I/O thread spun for 1000 iterations
>>>>> main-loop: WARNING: I/O thread spun for 1000 iterations
>>>>> main-loop: WARNING: I/O thread spun for 1000 iterations
>>>>> [vmxnet3][WR][vmxnet3_peer_has_vnet_hdr]: Peer has no virtio
>>>>> extension. Task offloads will be emulated.
>>>>> make: *** [check-qtest-i386] Error 1
>>>>>
>>>>> It might be an intermittent test fail from an earlier IDE pull?
>>>> Yep, intermittent. Second run of the tests passed...
>>> and repro'd on current master, so definitely not the fault of anything
>>> in this pull.
>> while QTEST_QEMU_BINARY=i386-softmmu/qemu-system-i386
>> QTEST_QEMU_IMG=qemu-img gtester -k --verbose -m=quick tests/ide-test ;
>> do true; done
>>
>> will eventually hang on /i386/ide/cdrom/pio_large (may take a fair
>> number of loops; when it works fine the test does not take an
>> appreciable amount of time). After a long time the timeout in the
>> test kicks in and the assert happens.
> cc'ing in Peter Lieven given his recent IDE buffering changes.
I wonder if there is a glitch in the PIO implementation of test-ide.c. As far
as I understand the specs
it is not allowed to read data while the BSY flag is set. With the following
change to the test-ide script
the test does not race:
diff --git a/tests/ide-test.c b/tests/ide-test.c
index d1014bb..ab0489e 100644
--- a/tests/ide-test.c
+++ b/tests/ide-test.c
@@ -728,6 +728,7 @@ static void cdrom_pio_impl(int nblocks)
for (i = 0; i < DIV_ROUND_UP(rxsize, limit); i++) {
size_t offset = i * (limit / 2);
size_t rem = (rxsize / 2) - offset;
+ ide_wait_clear(BSY);
for (j = 0; j < MIN((limit / 2), rem); j++) {
rx[offset + j] = le16_to_cpu(inw(IDE_BASE + reg_data));
}
Note: in the old sync version of the ATAPI PIO implementation this could not
happen.
Peter
