[yocto] prelink-cross project sync update
Mark Hatle
mark.hatle at windriver.com
Wed Oct 26 10:41:28 PDT 2011
(I'm going to try to send out these quick blurbs as prelink-cross is synced to
the upstream. Upstream changes are fairly infrequent, so this will be a low
volume set of emails. Most emails in the future will be shorted then this one.)
prelink-cross has been synced up to the latest upstream prelink changes.
The latest changes include:
2011-10-12 Jakub Jelinek <jakub at redhat.com>
* src/main.c (layout_page_size): New variable.
(OPT_LAYOUT_PAGE_SIZE): Define.
(options): Add --layout-page-size option.
(set_default_layout_page_size): New function in i?86/x86_64
resp. empty macro on other architectures.
(main): Call it.
* src/arch-i386.c (i386_layout_libs_init): Fail if
l->max_page_size is larger than 2MB and exec_shield is true.
Compute reg0s as REG0S rounded up to l->max_page_size
boundary and use it instead of REG0S.
(i386_layout_libs_pre): Compute reg0s as REG0S rounded up to
l->max_page_size boundary and use it instead of REG0S.
Fix up a pasto in computation of virt parameter for
third i386_find_free_addr call.
* src/prelink.h (layout_page_size): New decl.
* src/layout.c (layout_libs): Set max_page_size and
l.max_page_size to maximum of plarch->max_page_size
and layout_page_size. Fail if mmap_base or mmap_end
aren't max_page_size aligned.
* src/arch-ppc.c (ppc_layout_libs_pre): Give up if
l->max_page_size is larger than 64KB.
In essence what this does is adds a new option "--layout-page-size" that allows
someone to override the "max_page_size" value when necessary. (It's not a true
override as the max_page_size is set to either the existing value or
layout-page-size whichever is larger.)
The upstream change also added dynamic run-time detection for the AMD Bulldozer
platform, setting the default to 32k. Since run-time detection is contrary to
the existing prelink-cross usage, the code has been disabled.
As usual, the upstream version is in the "master" branch, and the cross version
is in the "cross_prelink" branch.
--Mark
More information about the yocto
mailing list