summaryrefslogtreecommitdiffstats
path: root/meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch
diff options
context:
space:
mode:
authorKhem Raj <raj.khem@gmail.com>2016-03-22 17:06:07 +0000
committerRichard Purdie <richard.purdie@linuxfoundation.org>2016-03-25 10:29:16 +0000
commit4fe64d79a08cc08f3322546136e35a080aa5458b (patch)
tree8480bd2ba8ba19b1d3b3d870d6f2f41f98515aec /meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch
parent4aa08b897628ef43adc9c3aa86bdf32499c328c1 (diff)
downloadpoky-4fe64d79a08cc08f3322546136e35a080aa5458b.tar.gz
libunwind: Fix build with fstack-protector on musl
libunwind makery inserts -nostdlib during linking which fails the build on musl when security flags are enabled since it remove ssp from linking, so add them explicitly to SECURITY_LDFLAGS disable tests for musl targets, tests use obsolete posix APIs e.g. getcontext patchout x86_local_resume() on x86, gets a working linunwind on x86, it seems that it wont work even in glibc case but lets leave it as it is for glibc and apply the patch only for musl (From OE-Core rev: c8ac9d483f6e1cfca82dad8cf3e0745935e96214) Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Diffstat (limited to 'meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch')
-rw-r--r--meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch54
1 files changed, 54 insertions, 0 deletions
diff --git a/meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch b/meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch
new file mode 100644
index 0000000000..371013aaaa
--- /dev/null
+++ b/meta/recipes-support/libunwind/libunwind-1.1/0001-x86-Stub-out-x86_local_resume.patch
@@ -0,0 +1,54 @@
1From f5df01655a4b76d4fe415747de581d94ac593e6a Mon Sep 17 00:00:00 2001
2From: Khem Raj <raj.khem@gmail.com>
3Date: Tue, 22 Mar 2016 16:19:29 +0000
4Subject: [PATCH] x86: Stub out x86_local_resume()
5
6its purpose seems
7to be unwinding across signal handler boundaries, which cannot happen
8in correct programs anyway. Replacing the whole function with
9something like *(volatile char *)0=0; (i.e. crash), gets a working
10libunwind
11
12Signed-off-by: Khem Raj <raj.khem@gmail.com>
13---
14Upstream-Status: Pending
15
16 src/x86/Gos-linux.c | 22 +---------------------
17 1 file changed, 1 insertion(+), 21 deletions(-)
18
19diff --git a/src/x86/Gos-linux.c b/src/x86/Gos-linux.c
20index 31f83ba..3aaa34e 100644
21--- a/src/x86/Gos-linux.c
22+++ b/src/x86/Gos-linux.c
23@@ -281,27 +281,7 @@ x86_r_uc_addr (ucontext_t *uc, int reg)
24 HIDDEN int
25 x86_local_resume (unw_addr_space_t as, unw_cursor_t *cursor, void *arg)
26 {
27- struct cursor *c = (struct cursor *) cursor;
28- ucontext_t *uc = c->uc;
29-
30- /* Ensure c->pi is up-to-date. On x86, it's relatively common to be
31- missing DWARF unwind info. We don't want to fail in that case,
32- because the frame-chain still would let us do a backtrace at
33- least. */
34- dwarf_make_proc_info (&c->dwarf);
35-
36- if (unlikely (c->sigcontext_format != X86_SCF_NONE))
37- {
38- struct sigcontext *sc = (struct sigcontext *) c->sigcontext_addr;
39-
40- Debug (8, "resuming at ip=%x via sigreturn(%p)\n", c->dwarf.ip, sc);
41- sigreturn (sc);
42- }
43- else
44- {
45- Debug (8, "resuming at ip=%x via setcontext()\n", c->dwarf.ip);
46- setcontext (uc);
47- }
48+ *(volatile char *)0=0;
49 return -UNW_EINVAL;
50 }
51 #endif
52--
531.8.3.1
54