13534 2019-09-17 06:09:09 +0000 json-c: bitbake hang in recent AMD cpu. 2019-10-13 19:38:33 +0000 1 1 7 Build System, Metadata & Runtime OE-Core devtools / tool chain 2.6.4 x86 Multiple CLOSED FIXED Medium+ normal 2.8 M4 1 cassian.adr bunk akuster bunk meta.mr.watcher meta.watcher randy.macleod --- --- No (bug/feature does not impact docs) oldest_to_newest 84990 0 cassian.adr 2019-09-17 06:09:09 +0000 I am using recent AMD cpu, AMD Ryzen 9 3900X. When I bitbake (it'a AGL project) some recipes used 100% on build in one core and never finished. Same happened ubuntu_18.04, ubuntu_16.04 and debian_sid host. I found it hang in `json_c_get_random_seed()` and, can workaround it by comment out --enable-rdrand in `poky/meta/recipes-devtools/json-c/json-c_0.13.1.bb`: ... #EXTRA_OECONF = "--enable-rdrand" ... It makes json-c use `/dev/urandom` instead of RDRAND to get random seed. I thinks it because something about recent AMD cpu's RDRAND instruction but not know detail. Hope, this report helps someone in same problem. 85013 1 randy.macleod 2019-09-19 14:41:49 +0000 Can you reproduce the issue on a non-AMD builder? Also we'd like to know more about the exact branch and commit of oe-core used? Does it happen on master, there was a recent update to openssl that may be related. Any additional info would be appreciated or actually needed to make progress. 85031 2 cassian.adr 2019-09-20 00:35:33 +0000 It's not reproduced in Intel(R) Core(TM) i7-4770 CPU builder. Current tags are `thud-20.0.2` and `yocto-2.6.2` in remote, `https://git.yoctoproject.org/git/poky` poky's `master` branch also has same EXTRA_OECONF, `--enable-rdrand` in `json-c_0.13.1.bb` so, will reproduce same problem I think. It may hang from here which use resigiter directly with asm code in json-c * https://github.com/json-c/json-c/blob/eae040a84a479ccad1d1c48314345c51ecf1a4a4/random_seed.c#L72 or Think it's not related with openssl updates. Thanks. 85033 3 bunk 2019-09-20 07:47:01 +0000 (In reply to comment #1) > Can you reproduce the issue on a non-AMD builder? > Also we'd like to know more about the exact branch and commit of oe-core > used? > Does it happen on master, there was a recent update to openssl that may be > related. Any additional info would be appreciated or actually needed to make > progress. This is not related to OpenSSL. RDRAND on latest Ryzen hardware has known problems, AFAIR there should be BIOS updates available that fix/workaround the problem. 85034 4 bunk 2019-09-20 08:10:03 +0000 I'll send a patch to remove the --enable-rdrand. 85337 5 akuster 2019-10-13 19:38:22 +0000 in warrior: https://git.openembedded.org/openembedded-core/commit/?h=warrior&id=07b19cd9ef127ec87f727e154eda62272a828275 Thud: https://git.openembedded.org/openembedded-core/commit/?h=thud&id=fad633eb5c464d4e2a984b9259625bcd150ee357 85338 6 akuster 2019-10-13 19:38:33 +0000 closing