public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
From: "redi at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org> To: gcc-bugs@gcc.gnu.org Subject: [Bug libstdc++/100444] std::random_device isn't random on AMD Date: Thu, 06 May 2021 11:25:36 +0000 [thread overview] Message-ID: <bug-100444-4-NYwyv1t4Eq@http.gcc.gnu.org/bugzilla/> (raw) In-Reply-To: <bug-100444-4@http.gcc.gnu.org/bugzilla/> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100444 --- Comment #8 from Jonathan Wakely <redi at gcc dot gnu.org> --- g:b0c0d878a8b5bf39dbea4c192fed26d340524439 enabled RDRAND and RDSEED for AMD chips, where previously we'd only used RDRAND and only for Intel (which caused one particular idiot to complain that we "hate AMD"). I think we originally didn't do it for AMD because of an older (but similar) bug: https://bugzilla.kernel.org/show_bug.cgi?id=85911 N.B. If you know you have a broken chip and can change the source you can select a different source, e.g. std::random_device rd("/dev/urandom"). I think this should work: --- a/libstdc++-v3/src/c++11/random.cc +++ b/libstdc++-v3/src/c++11/random.cc @@ -101,6 +101,19 @@ namespace std _GLIBCXX_VISIBILITY(default) return val; } + + bool + __attribute__ ((target("rdrnd"))) + __x86_rdrand_is_usable() + { + for (int borkcount = 0; borkcount < 10; ++borkcount) + { + // AMD Ryzen 3000 bug, see PR libstdc++/100444 + if (__x86_rdrand(nullptr) != 0xffffffff) + return true; + } + return false; + } #endif #if USE_RDSEED @@ -271,7 +284,7 @@ namespace std _GLIBCXX_VISIBILITY(default) #ifdef USE_RDRAND // CPUID.01H:ECX.RDRAND[bit 30] __cpuid(1, eax, ebx, ecx, edx); - if (ecx & bit_RDRND) + if (ecx & bit_RDRND && __x86_rdrand_is_usable()) { _M_func = &__x86_rdseed_rdrand; return; @@ -297,8 +310,13 @@ namespace std _GLIBCXX_VISIBILITY(default) __cpuid(1, eax, ebx, ecx, edx); if (ecx & bit_RDRND) { - _M_func = &__x86_rdrand; - return; + if (__x86_rdrand_is_usable()) + { + _M_func = &__x86_rdrand; + return; + } + else if (which == rdrand) + __throw_runtime_error(__N("random_device: RDRAND is buggy")); } } } If you ask for "rdrand" specifically you'll get an exception telling you it's buggy. If you just let the library choose for you, then it will skip rdrand and pick the next alternative, which will be "/dev/urandom" for most systems.
next prev parent reply other threads:[~2021-05-06 11:25 UTC|newest] Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-05-06 7:45 [Bug libstdc++/100444] New: " ecree429 at virginmedia dot com 2021-05-06 7:52 ` [Bug libstdc++/100444] " rguenth at gcc dot gnu.org 2021-05-06 8:48 ` ecree429 at virginmedia dot com 2021-05-06 8:49 ` pinskia at gcc dot gnu.org 2021-05-06 8:58 ` rguenth at gcc dot gnu.org 2021-05-06 10:05 ` redi at gcc dot gnu.org 2021-05-06 10:22 ` redi at gcc dot gnu.org 2021-05-06 10:51 ` ecree429 at virginmedia dot com 2021-05-06 11:25 ` redi at gcc dot gnu.org [this message] 2021-05-06 11:39 ` rguenth at gcc dot gnu.org 2021-05-06 11:41 ` redi at gcc dot gnu.org 2021-05-10 20:20 ` redi at gcc dot gnu.org 2021-05-13 18:46 ` ecree429 at virginmedia dot com 2021-05-14 9:23 ` redi at gcc dot gnu.org 2021-09-25 20:07 ` redi at gcc dot gnu.org 2021-10-04 12:20 ` redi at gcc dot gnu.org
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=bug-100444-4-NYwyv1t4Eq@http.gcc.gnu.org/bugzilla/ \ --to=gcc-bugzilla@gcc.gnu.org \ --cc=gcc-bugs@gcc.gnu.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).