Date: Tue, 13 Mar 2012 13:27:15 -0700 From: Oleksandr Tymoshenko <gonzo@freebsd.org> To: Marcel Moolenaar <marcel@xcllnt.net> Cc: svn-src-head@FreeBSD.org, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org Subject: Re: svn commit: r232582 - head/lib/libc/gen Message-ID: <4F5FADA3.4070902@freebsd.org> In-Reply-To: <DCC87BC2-A305-42A8-A861-17710A796B56@xcllnt.net> References: <201203060342.q263gs7g018067@svn.freebsd.org> <DCC87BC2-A305-42A8-A861-17710A796B56@xcllnt.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 13/03/2012 9:21 AM, Marcel Moolenaar wrote: > > On Mar 5, 2012, at 7:42 PM, Oleksandr Tymoshenko wrote: >> @@ -308,6 +308,13 @@ _init_tls() >> } >> } >> >> +#ifdef TLS_VARIANT_I >> + /* >> + * tls_static_space should include space for TLS structure >> + */ >> + tls_static_space += TLS_TCB_SIZE; >> +#endif >> + > > How does this change behaviour on ia64 or powerpc? I believe it doesn't. nathanw@ performed some sanity tests on powerpc recently, and they passed. ia64 has not been tested. This issue manifested itself only with statically linked binaries. Juli presented some corner case test for MIPS that caused crash: binary compiled with "-static -pthread" without any actual thread- specific variables that tries to get TP pointer. This test case leads to crash and I used it for tracking down the root cause. For binaries with actual TLS data this issue would manifest itself as memory corruption.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4F5FADA3.4070902>