On Mon, Jun 08, 2015 at 02:26:03PM +0200, Gerd Hoffmann wrote: > > Hi. I'm afraid this fails make check for me: > > > > GTESTER check-qtest-ppc > > qemu-system-ppc: > > /home/petmay01/linaro/qemu-for-merges/hw/nvram/fw_cfg.c:426: > > fw_cfg_add_bytes_read_callback: Assertion `s->entries[arch][key].data > > == ((void *)0)' failed. > > Broken pipe > > GTester: last random seed: R02Sd9273ba718a6b28ff4381eed659a0cb4 > > qemu-system-ppc: > > /home/petmay01/linaro/qemu-for-merges/hw/nvram/fw_cfg.c:426: > > fw_cfg_add_bytes_read_callback: Assertion `s->entries[arch][key].data > > == ((void *)0)' failed. > > Broken pipe > > GTester: last random seed: R02S59d67a130b823e6e6d04a41eb4b168c7 > > Reproducer: "qemu-system-ppc -M g3beige -boot once=d,order=c". > Culprit: Patch #3 (fw_cfg: prevent selector key conflict). > Looks like FW_CFG_BOOT_DEVICE is set twice for macs. > > Gabriel? To me this looks like the assert found an actual bug ...
I think it did -- a (small) memory leak in the fw_cfg_boot_set() callback. I'm sending out two patches which should go in before the current set in this pull request (to maintain "bisectability"). Thanks, --Gabriel
