x86/vdso: Remove obsolete "fake section table" reservation
authorSean Christopherson <sean.j.christopherson@intel.com>
Tue, 4 Dec 2018 21:25:57 +0000 (13:25 -0800)
committerIngo Molnar <mingo@kernel.org>
Wed, 5 Dec 2018 07:58:11 +0000 (08:58 +0100)
At one point the vDSO image was manually stripped down by vdso2c in an
attempt to minimize the size of the image mapped into userspace.  Part
of that stripping process involved building a fake section table so as
not to break userspace processes that parse the section table.  Memory
for the fake section table was reserved in the .rodata section so that
vdso2c could simply copy the entire PT_LOAD segment into the userspace
image after building the fake table.

Eventually, the entire fake section table approach was dropped in favor
of stripping the vdso "the old fashioned way", i.e. via objdump -S.
But, the reservation in .rodata for the fake table was left behind.
Remove the reserveration along with a few other related defines and
section entries.

Removing the fake section table placeholder zaps a whopping 0x340 bytes
from the 64-bit vDSO image, which drops the current image's size to
under 4k, i.e. reduces the effective size of the userspace vDSO mapping
by a full page.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Acked-by: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: da861e18eccc ("x86, vdso: Get rid of the fake section mechanism")
Link: http://lkml.kernel.org/r/20181204212600.28090-2-sean.j.christopherson@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/entry/vdso/vdso-layout.lds.S
arch/x86/entry/vdso/vdso2c.c

index acfd5ba..0cedc90 100644 (file)
@@ -7,16 +7,6 @@
  * This script controls its layout.
  */
 
-#if defined(BUILD_VDSO64)
-# define SHDR_SIZE 64
-#elif defined(BUILD_VDSO32) || defined(BUILD_VDSOX32)
-# define SHDR_SIZE 40
-#else
-# error unknown VDSO target
-#endif
-
-#define NUM_FAKE_SHDRS 13
-
 SECTIONS
 {
        /*
@@ -60,20 +50,8 @@ SECTIONS
                *(.bss*)
                *(.dynbss*)
                *(.gnu.linkonce.b.*)
-
-               /*
-                * Ideally this would live in a C file, but that won't
-                * work cleanly for x32 until we start building the x32
-                * C code using an x32 toolchain.
-                */
-               VDSO_FAKE_SECTION_TABLE_START = .;
-               . = . + NUM_FAKE_SHDRS * SHDR_SIZE;
-               VDSO_FAKE_SECTION_TABLE_END = .;
        }                                               :text
 
-       .fake_shstrtab  : { *(.fake_shstrtab) }         :text
-
-
        .note           : { *(.note.*) }                :text   :note
 
        .eh_frame_hdr   : { *(.eh_frame_hdr) }          :text   :eh_frame_hdr
index 4674f58..8e470b0 100644 (file)
@@ -76,8 +76,6 @@ enum {
        sym_hpet_page,
        sym_pvclock_page,
        sym_hvclock_page,
-       sym_VDSO_FAKE_SECTION_TABLE_START,
-       sym_VDSO_FAKE_SECTION_TABLE_END,
 };
 
 const int special_pages[] = {
@@ -98,12 +96,6 @@ struct vdso_sym required_syms[] = {
        [sym_hpet_page] = {"hpet_page", true},
        [sym_pvclock_page] = {"pvclock_page", true},
        [sym_hvclock_page] = {"hvclock_page", true},
-       [sym_VDSO_FAKE_SECTION_TABLE_START] = {
-               "VDSO_FAKE_SECTION_TABLE_START", false
-       },
-       [sym_VDSO_FAKE_SECTION_TABLE_END] = {
-               "VDSO_FAKE_SECTION_TABLE_END", false
-       },
        {"VDSO32_NOTE_MASK", true},
        {"__kernel_vsyscall", true},
        {"__kernel_sigreturn", true},