Skip to contentSkip to conversation

syscalls/mprotect: align exec_func to 64 bytes

PATCH Superseded Jan Stancek 3 participants 5 messages Feb 11, 2019 – Feb 25, 2019

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 syscalls/mprotect: align exec_func to 64 bytes

exec_func() is dummy/empty function. Try to align it so we don't need to worry about copying 2 pages. But also check that compiler aligned it and there's sufficient space between start of func_exec and end of page.

This patch also removes copy_sz, which is now replaced with page_sz.

---
 testcases/kernel/syscalls/mprotect/Makefile     |  2 +
 testcases/kernel/syscalls/mprotect/mprotect04.c | 55 +++++++++++--------------
 2 files changed, 27 insertions(+), 30 deletions(-)
testcases/kernel/syscalls/mprotect/Makefile +2 -0
index bd617d806675..bc5c8bc10395 100644
@@ -20,4 +20,6 @@ top_srcdir		?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
+mprotect04: CFLAGS += -falign-functions=64
+
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/mprotect/mprotect04.c +25 -30

130 lines not on this page · raw message

Signature · 1 line
1.8.3.1
#1 lore raw

PATCH v2 1/1 syscalls/mprotect: align exec_func to 64 bytes

Two comments below. Otherwise, I'm fine with this. I reviewed this patch and found it to be working on an aarch64 platform.

Quoted 88 lines · Jan Stancek

exec_func() is dummy/empty function. Try to align it so we don't need to worry about copying 2 pages. But also check that compiler aligned it and there's sufficient space between start of func_exec and end of page.

This patch also removes copy_sz, which is now replaced with page_sz.

---
 testcases/kernel/syscalls/mprotect/Makefile     |  2 +
 testcases/kernel/syscalls/mprotect/mprotect04.c | 55 +++++++++++--------------
 2 files changed, 27 insertions(+), 30 deletions(-)
testcases/kernel/syscalls/mprotect/Makefile +2 -0
index bd617d806675..bc5c8bc10395 100644
@@ -20,4 +20,6 @@ top_srcdir            ?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
+mprotect04: CFLAGS += -falign-functions=64
+
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/mprotect/mprotect04.c +6 -10
index 60941a4220d5..0f7890dca03b 100644
@@ -55,7 +55,7 @@ int TST_TOTAL = ARRAY_SIZE(testfunc);
 
 static volatile int sig_caught;
 static sigjmp_buf env;
-static unsigned int copy_sz;
+static unsigned int page_sz;
 typedef void (*func_ptr_t)(void);
 
 int main(int ac, char **av)
@@ -88,7 +88,7 @@ static void setup(void)
 {
        tst_tmpdir();
        tst_sig(NOFORK, sighandler, cleanup);
-       copy_sz = getpagesize() * 2;
+       page_sz = getpagesize();
 
        TEST_PAUSE;
 }
@@ -96,12 +96,9 @@ static void setup(void)
 static void testfunc_protnone(void)
 {
        char *addr;
-       int page_sz;
 
        sig_caught = 0;
 
-       page_sz = getpagesize();
-
        addr = SAFE_MMAP(cleanup, 0, page_sz, PROT_READ | PROT_WRITE,
                                         MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 
@@ -133,7 +130,7 @@ static void testfunc_protnone(void)
 
 #ifdef __ia64__
 
-static char exec_func[] = {
+static char exec_func[] __attribute__ ((aligned (64))) = {
        0x11, 0x00, 0x00, 0x00, 0x01, 0x00, /* nop.m 0x0             */
        0x00, 0x00, 0x00, 0x02, 0x00, 0x80, /* nop.i 0x0             */
        0x08, 0x00, 0x84, 0x00,             /* br.ret.sptk.many b0;; */
@@ -210,42 +207,33 @@ typedef struct {
  * Copy page where &exec_func resides. Also try to copy subsequent page
  * in case exec_func is close to page boundary.
  */
-static void *get_func(void *mem)
+static void *get_func(void *mem, uintptr_t *func_page_offset)
 {
        uintptr_t page_sz = getpagesize();
        uintptr_t page_mask = ~(page_sz - 1);
-       uintptr_t func_page_offset;
        void *func_copy_start, *page_to_copy;
        void *mem_start = mem;
 
 #ifdef USE_FUNCTION_DESCRIPTORS
        func_descr_t *opd =  (func_descr_t *)&exec_func;
-       func_page_offset = (uintptr_t)opd->entry & (page_sz - 1);
-       func_copy_start = mem + func_page_offset;
+       *func_page_offset = (uintptr_t)opd->entry & (page_sz - 1);
+       func_copy_start = mem + *func_page_offset;

testcases/kernel/syscalls/mprotect/mprotect04.c typedef struct { · line 228

Quoted 10 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +4 -2
        page_to_copy = (void *)((uintptr_t)opd->entry & page_mask);
 #else
-       func_page_offset = (uintptr_t)&exec_func & (page_sz - 1);
-       func_copy_start = mem + func_page_offset;
+       *func_page_offset = (uintptr_t)&exec_func & (page_sz - 1);
+       func_copy_start = mem + *func_page_offset;
        page_to_copy = (void *)((uintptr_t)&exec_func & page_mask);
 #endif
+       tst_resm(TINFO, "exec_func: %p, page_to_copy: %p",
+               &exec_func, page_to_copy);

This was previously inside the body of the if statement right below. Not sure if it was intended to always print this information, but I'm fine either way.

Quoted 38 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +5 -14
        /* copy 1st page, if it's not present something is wrong */
-       if (!page_present(page_to_copy)) {
-               tst_resm(TINFO, "exec_func: %p, page_to_copy: %p\n",
-                       &exec_func, page_to_copy);
+       if (!page_present(page_to_copy))
                tst_brkm(TBROK, cleanup, "page_to_copy not present\n");
-       }
-       memcpy(mem, page_to_copy, page_sz);
 
-       /* copy 2nd page if possible */
-       mem += page_sz;
-       page_to_copy += page_sz;
-       if (page_present(page_to_copy))
-               memcpy(mem, page_to_copy, page_sz);
-       else
-               memset(mem, 0, page_sz);
+       memcpy(mem, page_to_copy, page_sz);
 
-       clear_cache(mem_start, copy_sz);
+       clear_cache(mem_start, page_sz);
 
        /* return pointer to area where copy of exec_func resides */
        return func_copy_start;
@@ -256,23 +244,30 @@ static void *get_func(void *mem)
 static void testfunc_protexec(void)
 {
        func_ptr_t func;
+       uintptr_t func_page_offset;
        void *p;
 
        sig_caught = 0;
 
-       p = SAFE_MMAP(cleanup, 0, copy_sz, PROT_READ | PROT_WRITE,
+       p = SAFE_MMAP(cleanup, 0, page_sz, PROT_READ | PROT_WRITE,
                 MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 
 #ifdef USE_FUNCTION_DESCRIPTORS

testcases/kernel/syscalls/mprotect/mprotect04.c static void *get_func(void *mem) · line 263

Quoted 10 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +3 -2
        func_descr_t opd;
-       opd.entry = (uintptr_t)get_func(p);
+       opd.entry = (uintptr_t)get_func(p, &func_page_offset);
        func = (func_ptr_t)&opd;
 #else
-       func = get_func(p);
+       func = get_func(p, &func_page_offset);
 #endif
 
+       if (func_page_offset + 64 >= page_sz) {

I'm wondering if this should be ">" not ">=". If the compiler decides to use the last 64 bytes in a page and locates the function at offset 0xfc0, then that's still ok: 0xfc0 + 0x40 == page_sz

Quoted 23 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +7 -3
+               SAFE_MUNMAP(cleanup, p, page_sz);
+               tst_brkm(TCONF, cleanup, "func too close to page boundary, "
+                       "maybe your compiler ignores -falign-functions?");
+       }
+
        /* Change the protection to PROT_EXEC. */
-       TEST(mprotect(p, copy_sz, PROT_EXEC));
+       TEST(mprotect(p, page_sz, PROT_EXEC));
 
        if (TEST_RETURN == -1) {
                tst_resm(TFAIL | TTERRNO, "mprotect failed");
@@ -294,7 +289,7 @@ static void testfunc_protexec(void)
                }
        }
 
-       SAFE_MUNMAP(cleanup, p, copy_sz);
+       SAFE_MUNMAP(cleanup, p, page_sz);
 }
 
 static void cleanup(void)
--

1.8.3.1

#2 lore raw

PATCH v2 1/1 syscalls/mprotect: align exec_func to 64 bytes

----- Original Message -----

Quoted 102 lines

Two comments below. Otherwise, I'm fine with this. I reviewed this patch and found it to be working on an aarch64 platform.

Quoted 98 lines · Jan Stancek

exec_func() is dummy/empty function. Try to align it so we don't need to worry about copying 2 pages. But also check that compiler aligned it and there's sufficient space between start of func_exec and end of page.

This patch also removes copy_sz, which is now replaced with page_sz.

---
 testcases/kernel/syscalls/mprotect/Makefile     |  2 +
 testcases/kernel/syscalls/mprotect/mprotect04.c | 55
 +++++++++++--------------
 2 files changed, 27 insertions(+), 30 deletions(-)
testcases/kernel/syscalls/mprotect/Makefile +2 -0
index bd617d806675..bc5c8bc10395 100644
@@ -20,4 +20,6 @@ top_srcdir            ?= ../../../..
 
 include $(top_srcdir)/include/mk/testcases.mk
 
+mprotect04: CFLAGS += -falign-functions=64
+
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/mprotect/mprotect04.c +8 -12
index 60941a4220d5..0f7890dca03b 100644
@@ -55,7 +55,7 @@ int TST_TOTAL = ARRAY_SIZE(testfunc);
 
 static volatile int sig_caught;
 static sigjmp_buf env;
-static unsigned int copy_sz;
+static unsigned int page_sz;
 typedef void (*func_ptr_t)(void);
 
 int main(int ac, char **av)
@@ -88,7 +88,7 @@ static void setup(void)
 {
        tst_tmpdir();
        tst_sig(NOFORK, sighandler, cleanup);
-       copy_sz = getpagesize() * 2;
+       page_sz = getpagesize();
 
        TEST_PAUSE;
 }
@@ -96,12 +96,9 @@ static void setup(void)
 static void testfunc_protnone(void)
 {
        char *addr;
-       int page_sz;
 
        sig_caught = 0;
 
-       page_sz = getpagesize();
-
        addr = SAFE_MMAP(cleanup, 0, page_sz, PROT_READ | PROT_WRITE,
                                         MAP_PRIVATE | MAP_ANONYMOUS, -1,
                                         0);
@@ -133,7 +130,7 @@ static void testfunc_protnone(void)
 
 #ifdef __ia64__
 
-static char exec_func[] = {
+static char exec_func[] __attribute__ ((aligned (64))) = {
        0x11, 0x00, 0x00, 0x00, 0x01, 0x00, /* nop.m 0x0             */
        0x00, 0x00, 0x00, 0x02, 0x00, 0x80, /* nop.i 0x0             */
        0x08, 0x00, 0x84, 0x00,             /* br.ret.sptk.many b0;; */
@@ -210,42 +207,33 @@ typedef struct {
  * Copy page where &exec_func resides. Also try to copy subsequent page
  * in case exec_func is close to page boundary.
  */
-static void *get_func(void *mem)
+static void *get_func(void *mem, uintptr_t *func_page_offset)
 {
        uintptr_t page_sz = getpagesize();
        uintptr_t page_mask = ~(page_sz - 1);
-       uintptr_t func_page_offset;
        void *func_copy_start, *page_to_copy;
        void *mem_start = mem;
 
 #ifdef USE_FUNCTION_DESCRIPTORS
        func_descr_t *opd =  (func_descr_t *)&exec_func;
-       func_page_offset = (uintptr_t)opd->entry & (page_sz - 1);
-       func_copy_start = mem + func_page_offset;
+       *func_page_offset = (uintptr_t)opd->entry & (page_sz - 1);
+       func_copy_start = mem + *func_page_offset;
        page_to_copy = (void *)((uintptr_t)opd->entry & page_mask);
 #else
-       func_page_offset = (uintptr_t)&exec_func & (page_sz - 1);
-       func_copy_start = mem + func_page_offset;
+       *func_page_offset = (uintptr_t)&exec_func & (page_sz - 1);
+       func_copy_start = mem + *func_page_offset;
Quoted 8 lines

testcases/kernel/syscalls/mprotect/mprotect04.c typedef struct { · line 228

Quoted 4 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +2 -0
        page_to_copy = (void *)((uintptr_t)&exec_func & page_mask);
 #endif
+       tst_resm(TINFO, "exec_func: %p, page_to_copy: %p",
+               &exec_func, page_to_copy);

This was previously inside the body of the if statement right below. Not sure if it was intended to always print this information, but I'm fine either way.

This was intentional, if we get a failure it seemed like good data.

Quoted 48 lines
Quoted 48 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +7 -16
        /* copy 1st page, if it's not present something is wrong */
-       if (!page_present(page_to_copy)) {
-               tst_resm(TINFO, "exec_func: %p, page_to_copy: %p\n",
-                       &exec_func, page_to_copy);
+       if (!page_present(page_to_copy))
                tst_brkm(TBROK, cleanup, "page_to_copy not present\n");
-       }
-       memcpy(mem, page_to_copy, page_sz);
 
-       /* copy 2nd page if possible */
-       mem += page_sz;
-       page_to_copy += page_sz;
-       if (page_present(page_to_copy))
-               memcpy(mem, page_to_copy, page_sz);
-       else
-               memset(mem, 0, page_sz);
+       memcpy(mem, page_to_copy, page_sz);
 
-       clear_cache(mem_start, copy_sz);
+       clear_cache(mem_start, page_sz);
 
        /* return pointer to area where copy of exec_func resides */
        return func_copy_start;
@@ -256,23 +244,30 @@ static void *get_func(void *mem)
 static void testfunc_protexec(void)
 {
        func_ptr_t func;
+       uintptr_t func_page_offset;
        void *p;
 
        sig_caught = 0;
 
-       p = SAFE_MMAP(cleanup, 0, copy_sz, PROT_READ | PROT_WRITE,
+       p = SAFE_MMAP(cleanup, 0, page_sz, PROT_READ | PROT_WRITE,
                 MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 
 #ifdef USE_FUNCTION_DESCRIPTORS
        func_descr_t opd;
-       opd.entry = (uintptr_t)get_func(p);
+       opd.entry = (uintptr_t)get_func(p, &func_page_offset);
        func = (func_ptr_t)&opd;
 #else
-       func = get_func(p);
+       func = get_func(p, &func_page_offset);
 #endif
 
Quoted 5 lines

testcases/kernel/syscalls/mprotect/mprotect04.c static void *get_func(void *mem) · line 263

Quoted 1 line
testcases/kernel/syscalls/mprotect/mprotect04.c +1 -0
+       if (func_page_offset + 64 >= page_sz) {

I'm wondering if this should be ">" not ">=". If the compiler decides to use the last 64 bytes in a page and locates the function at offset 0xfc0, then that's still ok: 0xfc0 + 0x40 == page_sz

Agreed, I'll make it '>'.

Quoted 26 lines
Quoted 25 lines
testcases/kernel/syscalls/mprotect/mprotect04.c +2 -0
+               SAFE_MUNMAP(cleanup, p, page_sz);
+               tst_brkm(TCONF, cleanup, "func too close to page boundary,

"

testcases/kernel/syscalls/mprotect/mprotect04.c +5 -3
+                       "maybe your compiler ignores -falign-functions?");
+       }
+
        /* Change the protection to PROT_EXEC. */
-       TEST(mprotect(p, copy_sz, PROT_EXEC));
+       TEST(mprotect(p, page_sz, PROT_EXEC));
 
        if (TEST_RETURN == -1) {
                tst_resm(TFAIL | TTERRNO, "mprotect failed");
@@ -294,7 +289,7 @@ static void testfunc_protexec(void)
                }
        }
 
-       SAFE_MUNMAP(cleanup, p, copy_sz);
+       SAFE_MUNMAP(cleanup, p, page_sz);
 }
 
 static void cleanup(void)
--

1.8.3.1

#3 lore raw

PATCH v2 1/1 syscalls/mprotect: align exec_func to 64 bytes

In reply to Jan Stancek
Quoted 9 lines · Jan Stancek

exec_func() is dummy/empty function. Try to align it so we don't need to worry about copying 2 pages. But also check that compiler aligned it and there's sufficient space between start of func_exec and end of page.

This patch also removes copy_sz, which is now replaced with page_sz.

Quoted 2 lines

+mprotect04: CFLAGS += -falign-functions=64

One of my concern was this option maybe not supported by kind of compiler and probably broken in the compiling phase. So I just submitted these two(another is Daniel's) patches to Travis CI, the result looks good and no compiler error found there.
https://travis-ci.com/wangli5665/ltp/builds/101429714

Beside that, I also manually tried that with old gcc(gcc version 4.1.2) and verify that works fine.

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20190219/73dc6da9/attachment.html>
#4 lore raw

PATCH v2 1/1 syscalls/mprotect: align exec_func to 64 bytes

----- Original Message -----

Quoted 7 lines · Jan Stancek
Quoted 6 lines

exec_func() is dummy/empty function. Try to align it so we don't need to worry about copying 2 pages. But also check that compiler aligned it and there's sufficient space between start of func_exec and end of page.

Quoted 6 lines
Quoted 4 lines

This patch also removes copy_sz, which is now replaced with page_sz.

Pushed.

Regards,
Jan

Quoted 18 lines
Quoted 4 lines

+mprotect04: CFLAGS += -falign-functions=64

One of my concern was this option maybe not supported by kind of compiler and probably broken in the compiling phase. So I just submitted these two(another is Daniel's) patches to Travis CI, the result looks good and no compiler error found there.
https://travis-ci.com/wangli5665/ltp/builds/101429714

Beside that, I also manually tried that with old gcc(gcc version 4.1.2) and verify that works fine.

--
Regards,
Li Wang