I am proposing the following changes to the powerpc linker scripts (both
B and BDT). This change will align the end of the BSS to a 256M
boundary. Due to the restrictions on powerpc allow only one page size
per 256M chunk of virtual address space, it is possible to have spurious
remapping and/or malloc failures with some applications. We're seeing
it in the case where an application requests segment remapping, but the
loader mapped a library into the same 256 segment as the BSS. In this
case we can unmap the original BSS, but are not able to create it in
large pages due to the library in the way. Aligning the bss in this way
should not cause any harm. It may cause us to consume a few more huge
pages in some cases. This side effect is fixable in the future.
Does anyone see any problems with doing this?
Signed-Off-By: Adam Litke <[EMAIL PROTECTED]>
diff --git a/ldscripts/elf32ppclinux.xB b/ldscripts/elf32ppclinux.xB
index 8d5cf4d..e45e0d7 100644
--- a/ldscripts/elf32ppclinux.xB
+++ b/ldscripts/elf32ppclinux.xB
@@ -205,10 +205,8 @@ SECTIONS
.bss section disappears because there are no input sections.
FIXME: Why do we need it? When there is no .bss section, we don't
pad the .data section. */
- . = ALIGN(. != 0 ? 32 / 8 : 1);
+ . = ALIGN(256*1204*1024);
} :htlb
- . = ALIGN(32 / 8);
- . = ALIGN(32 / 8);
_end = .;
PROVIDE (end = .);
/* Stabs debugging sections. */
diff --git a/ldscripts/elf32ppclinux.xBDT b/ldscripts/elf32ppclinux.xBDT
index 8794edb..2734ebe 100644
--- a/ldscripts/elf32ppclinux.xBDT
+++ b/ldscripts/elf32ppclinux.xBDT
@@ -199,10 +199,8 @@ SECTIONS
.bss section disappears because there are no input sections.
FIXME: Why do we need it? When there is no .bss section, we don't
pad the .data section. */
- . = ALIGN(. != 0 ? 32 / 8 : 1);
+ . = ALIGN(256*1024*1024);
} :hdata
- . = ALIGN(32 / 8);
- . = ALIGN(32 / 8);
_end = .;
PROVIDE (end = .);
/* Stabs debugging sections. */
diff --git a/ldscripts/elf64ppc.xB b/ldscripts/elf64ppc.xB
index 073735f..48904f3 100644
--- a/ldscripts/elf64ppc.xB
+++ b/ldscripts/elf64ppc.xB
@@ -199,6 +199,7 @@ SECTIONS
*(.dynbss)
*(.bss .bss.* .gnu.linkonce.b.*)
*(COMMON)
+ . = ALIGN(256*1024*1024);
}
_end = .;
PROVIDE (end = .);
diff --git a/ldscripts/elf64ppc.xBDT b/ldscripts/elf64ppc.xBDT
index fc99882..69065ea 100644
--- a/ldscripts/elf64ppc.xBDT
+++ b/ldscripts/elf64ppc.xBDT
@@ -192,10 +192,8 @@ SECTIONS
.bss section disappears because there are no input sections.
FIXME: Why do we need it? When there is no .bss section, we don't
pad the .data section. */
- . = ALIGN(. != 0 ? 64 / 8 : 1);
+ . = ALIGN(256*1024*1024);
} :hdata
- . = ALIGN(64 / 8);
- . = ALIGN(64 / 8);
_end = .;
PROVIDE (end = .);
/*. = DATA_SEGMENT_END (.);*/
--
Adam Litke - (agl at us.ibm.com)
IBM Linux Technology Center
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Libhugetlbfs-devel mailing list
Libhugetlbfs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
