Test brk VMA code path
v2 cover Test brk VMA code path
The brk system call uses a slightly different code path through the kernel to expand/contract across VMAs. This new test is written to force the VMA to create at least two new entries and shrink back across VMA boundaries.
V2 changes:
- White space fixes.
- Removed unnecessary include
- Added tst_res in failure paths
Liam R. Howlett (1):
brk02: Add test for removing more than one VMA
testcases/kernel/syscalls/brk/brk02.c | 54 +++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) create mode 100644 testcases/kernel/syscalls/brk/brk02.c
Signature · 1 line
2.30.0
PATCH v2 1/1 brk02: Add test for removing more than one VMA
When brk expands, it attempts to expand a VMA. This expansion will succeed depending on the anonymous VMA chain and if the vma flags are compatible. This test expands brk() then calls mprotect to ensure the next brk call will create a new VMA, then it calls brk a final time to restore the first brk address. The test is the final brk call which will remove more than an entire VMA from the vm area.
--- testcases/kernel/syscalls/brk/brk02.c | 54 +++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) create mode 100644 testcases/kernel/syscalls/brk/brk02.c
testcases/kernel/syscalls/brk/brk02.c new file +54 -0
@@ -0,0 +1,54 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* +* Copyright (c) 2021 Liam R. Howlett <[email protected]> +* +* Expand the brk by at least 2 pages to ensure there is a newly created VMA +* and not expanding the original due to multiple anon pages. mprotect that +* new VMA then brk back to the original address therefore causing a munmap of +* at least one full VMA. +*/ + +#include <unistd.h> +#include <sys/mman.h> +#include "tst_test.h" + +void brk_down_vmas (void ) +{ +void *brk_addr =sbrk (0 ); +unsigned long page_size =getpagesize (); +void *addr = brk_addr + page_size; + +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by page size." ); +return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by 2x page size." ); +return ; + } + +if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk after mprotect." ); +return ; + } + +if (brk (brk_addr)) { +tst_res (TFAIL ,"Cannot restore brk to start address." ); +return ; + } + +tst_res (TPASS ,"munmap at least two VMAs of brk() passed." ); +} + +static struct tst_test test = { + .test_all = brk_down_vmas, +};
Signature · 1 line
2.30.0
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Liam,
Quoted 19 lines
When brk expands, it attempts to expand a VMA. This expansion will succeed depending on the anonymous VMA chain and if the vma flags are compatible. This test expands
brk()then calls mprotect to ensure the next brk call will create a new VMA, then it calls brk a final time to restore the first brk address. The test is the final brk call which will remove more than an entire VMA from the vm area.Signed-off-by Liam R. Howlett <[email protected]>--- testcases/kernel/syscalls/brk/brk02.c | 54 +++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) create mode 100644 testcases/kernel/syscalls/brk/brk02.ctestcases/kernel/syscalls/brk/brk02.c new file +0
-0
testcases/kernel/syscalls/brk/brk02.c line 8
Quoted 8 lines
testcases/kernel/syscalls/brk/brk02.c new file +8
-0@@ -0,0 +1,54 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* +* Copyright (c) 2021 Liam R. Howlett <[email protected]> +* +* Expand the brk by at least 2 pages to ensure there is a newly created VMA +* and not expanding the original due to multiple anon pages. mprotect that +* new VMA then brk back to the original address therefore causing a munmap of +* at least one full VMA.
I'll put this as a docparse formatting (will show in our documentation):
/*\
* [DESCRIPTION]
* Expand brk() by at least 2 pages to ensure there is a newly created VMA
* and not expanding the original due to multiple anon pages. mprotect() that
* new VMA then brk() back to the original address therefore causing a munmap of
* at least one full VMA.
\*/
testcases/kernel/syscalls/brk/brk02.c line 17
Quoted 9 lines
testcases/kernel/syscalls/brk/brk02.c new file +9
-0+*/ + +#include <unistd.h> +#include <sys/mman.h> +#include "tst_test.h" + +void brk_down_vmas (void ) +{ +void *brk_addr =sbrk (0 );
Shouldn't there be a check?
if (brk_addr == (void *) -1) tst_brk(TBROK, "sbrk() failed");
testcases/kernel/syscalls/brk/brk02.c line 22
Quoted 5 lines
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+unsigned long page_size =getpagesize (); +void *addr = brk_addr + page_size; + +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by page size." );
nit: remove dot at the end.
Quoted 22 lines
testcases/kernel/syscalls/brk/brk02.c new file +22
-0+return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by 2x page size." ); +return ; + } + +if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk after mprotect." ); +return ; + } + +if (brk (brk_addr)) {
testcases/kernel/syscalls/brk/brk02.c line 54
Quoted 10 lines
testcases/kernel/syscalls/brk/brk02.c new file +10
-0+tst_res (TFAIL ,"Cannot restore brk to start address." ); +return ; + } + +tst_res (TPASS ,"munmap at least two VMAs of brk() passed." ); +} + +static struct tst_test test = { + .test_all = brk_down_vmas, +};
No need to repost if you agree with these changes below.
Kind regards,
Petr
kernel/syscalls/brk/brk02.c +18 -11
@@ -1,13 +1,16 @@// SPDX-License-Identifier: GPL-2.0-or-later /* * Copyright (c) 2021 Liam R. Howlett <[email protected]> -* -* Expand the brk by at least 2 pages to ensure there is a newly created VMA -* and not expanding the original due to multiple anon pages. mprotect that -* new VMA then brk back to the original address therefore causing a munmap of -* at least one full VMA. */ +/*\ +* [DESCRIPTION] +* Expand brk() by at least 2 pages to ensure there is a newly created VMA +* and not expanding the original due to multiple anon pages. mprotect() that +* new VMA then brk() back to the original address therefore causing a munmap of +* at least one full VMA. +\*/ +#include <unistd.h> #include <sys/mman.h> #include "tst_test.h"
@@ -15,38 +18,42 @@void brk_down_vmas (void ) {void *brk_addr =sbrk (0 ); + +if (brk_addr == (void *)-1 ) +tst_brk (TBROK ,"sbrk() failed" ); +unsigned long page_size =getpagesize ();void *addr = brk_addr + page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk by page size." ); +tst_res (TFAIL ,"Cannot expand brk() by page size" );return ; } addr += page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk by 2x page size." ); +tst_res (TFAIL ,"Cannot expand brk() by 2x page size" );return ; }if (mprotect (addr - page_size, page_size,PROT_READ |PROT_WRITE |PROT_EXEC )) { -tst_res (TFAIL ,"Cannot mprotect new VMA." ); +tst_res (TFAIL ,"Cannot mprotect new VMA" );return ; } addr += page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk after mprotect." ); +tst_res (TFAIL ,"Cannot expand brk() after mprotect" );return ; }if (brk (brk_addr)) { -tst_res (TFAIL ,"Cannot restore brk to start address." ); +tst_res (TFAIL ,"Cannot restore brk() to start address" );return ; } -tst_res (TPASS ,"munmap at least two VMAs of brk() passed." ); +tst_res (TPASS ,"munmap at least two VMAs of brk() passed" ); }static struct tst_test test = {
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hello Petr,
Quoted 15 lines · Petr Vorel
Hi Liam,
Quoted 12 lines
When brk expands, it attempts to expand a VMA. This expansion will succeed depending on the anonymous VMA chain and if the vma flags are compatible. This test expands
brk()then calls mprotect to ensure the next brk call will create a new VMA, then it calls brk a final time to restore the first brk address. The test is the final brk call which will remove more than an entire VMA from the vm area.Signed-off-by Liam R. Howlett <[email protected]>--- testcases/kernel/syscalls/brk/brk02.c | 54 +++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) create mode 100644 testcases/kernel/syscalls/brk/brk02.c
Quoted 15 lines
Quoted 6 lines
testcases/kernel/syscalls/brk/brk02.c new file +0
-0testcases/kernel/syscalls/brk/brk02.c line 8
Quoted 8 lines
testcases/kernel/syscalls/brk/brk02.c new file +8
-0@@ -0,0 +1,54 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* +* Copyright (c) 2021 Liam R. Howlett <[email protected]> +* +* Expand the brk by at least 2 pages to ensure there is a newly created VMA +* and not expanding the original due to multiple anon pages. mprotect that +* new VMA then brk back to the original address therefore causing a munmap of +* at least one full VMA. I'll put this as a docparse formatting (will show in our documentation):
Thanks
Quoted 21 lines
/*\
* [DESCRIPTION]
* Expandbrk()by at least 2 pages to ensure there is a newly created VMA
* and not expanding the original due to multiple anon pages.mprotect()that
* new VMA thenbrk()back to the original address therefore causing a munmap of
* at least one full VMA.
\*/testcases/kernel/syscalls/brk/brk02.c line 17
Quoted 9 lines
testcases/kernel/syscalls/brk/brk02.c new file +9
-0+*/ + +#include <unistd.h> +#include <sys/mman.h> +#include "tst_test.h" + +void brk_down_vmas (void ) +{ +void *brk_addr =sbrk (0 );Shouldn't there be a check?
Quoted 2 lines
if (brk_addr == (void *) -1) tst_brk(TBROK, "sbrk() failed");
Yes, you are correct. And -1 is the correct thing to check.
Quoted 7 lines
testcases/kernel/syscalls/brk/brk02.c line 22
Quoted 6 lines
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+unsigned long page_size =getpagesize (); +void *addr = brk_addr + page_size; + +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by page size." );nit: remove dot at the end.
Sounds good
Quoted 35 lines
Quoted 22 lines
testcases/kernel/syscalls/brk/brk02.c new file +22
-0+return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk by 2x page size." ); +return ; + } + +if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + } + + addr += page_size; +if (brk (addr)) { +tst_res (TFAIL ,"Cannot expand brk after mprotect." ); +return ; + } + +if (brk (brk_addr)) {testcases/kernel/syscalls/brk/brk02.c line 54
Quoted 10 lines
testcases/kernel/syscalls/brk/brk02.c new file +10
-0+tst_res (TFAIL ,"Cannot restore brk to start address." ); +return ; + } + +tst_res (TPASS ,"munmap at least two VMAs of brk() passed." ); +} + +static struct tst_test test = { + .test_all = brk_down_vmas, +};No need to repost if you agree with these changes below.
Quoted 2 lines
Kind regards,
Petr
The changes below look good.
Thanks,
Liam
Quoted 76 lines
kernel/syscalls/brk/brk02.c +18
-11@@ -1,13 +1,16 @@// SPDX-License-Identifier: GPL-2.0-or-later /* * Copyright (c) 2021 Liam R. Howlett <[email protected]> -* -* Expand the brk by at least 2 pages to ensure there is a newly created VMA -* and not expanding the original due to multiple anon pages. mprotect that -* new VMA then brk back to the original address therefore causing a munmap of -* at least one full VMA. */ +/*\ +* [DESCRIPTION] +* Expand brk() by at least 2 pages to ensure there is a newly created VMA +* and not expanding the original due to multiple anon pages. mprotect() that +* new VMA then brk() back to the original address therefore causing a munmap of +* at least one full VMA. +\*/ +#include <unistd.h> #include <sys/mman.h> #include "tst_test.h" @@ -15,38 +18,42 @@void brk_down_vmas (void ) {void *brk_addr =sbrk (0 ); + +if (brk_addr == (void *)-1 ) +tst_brk (TBROK ,"sbrk() failed" ); +unsigned long page_size =getpagesize ();void *addr = brk_addr + page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk by page size." ); +tst_res (TFAIL ,"Cannot expand brk() by page size" );return ; } addr += page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk by 2x page size." ); +tst_res (TFAIL ,"Cannot expand brk() by 2x page size" );return ; }if (mprotect (addr - page_size, page_size,PROT_READ |PROT_WRITE |PROT_EXEC )) { -tst_res (TFAIL ,"Cannot mprotect new VMA." ); +tst_res (TFAIL ,"Cannot mprotect new VMA" );return ; } addr += page_size;if (brk (addr)) { -tst_res (TFAIL ,"Cannot expand brk after mprotect." ); +tst_res (TFAIL ,"Cannot expand brk() after mprotect" );return ; }if (brk (brk_addr)) { -tst_res (TFAIL ,"Cannot restore brk to start address." ); +tst_res (TFAIL ,"Cannot restore brk() to start address" );return ; } -tst_res (TPASS ,"munmap at least two VMAs of brk() passed." ); +tst_res (TPASS ,"munmap at least two VMAs of brk() passed" ); }static struct tst_test test = {
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Liam,
Quoted 1 line
The changes below look good.
Good, patchset merged.
Thanks for your work.
Kind regards,
Petr
PATCH v2 1/1 brk02: Add test for removing more than one VMA
In reply to Liam HowlettHi Liam, Petr,
testcases/kernel/syscalls/brk/brk02.c line 36
Quoted 7 lines · Liam Howlett
...
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + }
We got permission denied here while performing the brk02 on x86_64/s390x(kernel-4.18~). After looking at the manual page of mprotect(), seems the access issue caused by PROT_EXEC.
"
POSIX says that the behavior of mprotect() is unspecified if it is applied to a region of memory that was not obtained via mmap(2).
...
Whether PROT_EXEC has any effect different from PROT_READ depends on processor architecture, kernel version, and process state. If READ_IMPLIES_EXEC is set in the process's personality flags (see personality(2)), specifying PROT_READ will implicitly add PROT_EXEC. "
# ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
After removing the PROT_EXEC:
# ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20210317/0b7d83c6/attachment.htm>
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hello Li,
Thank you for looking at this testcase.
Quoted 17 lines · Li Wang
Hi Liam, Petr,
testcases/kernel/syscalls/brk/brk02.c line 36
Quoted 7 lines · Liam Howlett
...
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + }We got permission denied here while performing the brk02 on x86_64/s390x(kernel-4.18~). After looking at the manual page of
mprotect(), seems the access issue caused byPROT_EXEC.
Quoted 9 lines
"
POSIX says that the behavior ofmprotect()is unspecified if it is applied to a region of memory that was not obtained via mmap(2).
...
WhetherPROT_EXEChas any effect different fromPROT_READdepends on processor architecture, kernel version, and process state. IfREAD_IMPLIES_EXECis set in the process's personality flags (see personality(2)), specifyingPROT_READwill implicitly addPROT_EXEC. "
Unforntunately, dropping the PROT_EXEC causes the VMA behaviour to change and does not test what this testcase is intended to test.
Removing the PROT_EXEC and the PROT_WRITE does test what is supposed to be tested. Can you verify that this works on the s390x?
Are you using real hardware to test this or can I use some sort of emulation to test on my side?
Thank you,
Liam
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Liam,
Quoted 34 lines · Liam Howlett
Hello Li,
Thank you for looking at this testcase.
Quoted 13 lines · Li Wang
Hi Liam, Petr,
testcases/kernel/syscalls/brk/brk02.c line 36
Quoted 7 lines · Liam Howlett
...
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + }Quoted 6 lines
We got permission denied here while performing the brk02 on x86_64/s390x(kernel-4.18~). After looking at the manual page of
mprotect(), seems the access issue caused byPROT_EXEC."
POSIX says that the behavior ofmprotect()is unspecified if it isapplied
Quoted 7 lines
to a region of memory that was not obtained via mmap(2).
...
WhetherPROT_EXEChas any effect different fromPROT_READdepends on processor architecture, kernel version, and process state. IfREAD_IMPLIES_EXECis set in the process's personality flags (see personality(2)), specifyingPROT_READwill implicitly addPROT_EXEC. "
Quoted 3 lines
Unforntunately, dropping the
PROT_EXECcauses the VMA behaviour to change and does not test what this testcase is intended to test.
Yes, I agree with this. But am not sure if Linux take some action on security
side to prevent setting PROT_EXEC permission arbitrary.
Quoted 4 lines
Removing the
PROT_EXECand thePROT_WRITEdoes test what is supposed to be tested. Can you verify that this works on the s390x?
Actually just removing the PROT_EXEC then the brk02 can be passed on all my platforms.
Quoted 4 lines
Are you using real hardware to test this or can I use some sort of emulation to test on my side?
It looks like easily to reproduce.
I get failed with both virtual system and bare metal, e.g. the first one is on my Fedora33-workstation. But the worth to say, brk02 passed with raspberry pi3 and pi4.
x86_64
-------------
# virt-what # echo $? 0 # uname -r 5.10.22-200.fc33.x86_64 # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
x86_64
-------------
# virt-what kvm # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
s390x
-------------
# uname -r 5.12.0-rc4.vdso+ # virt-what ibm_systemz ibm_systemz-zvm # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
armv7l -- raspberry-pi3
-----------------------------
# uname -r 5.4.96-v7.1.el7 # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed armv7l -- raspberry-pi4 ----------------------------- # uname -rm 5.10.17-v7l+ armv7l # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20210325/5fe42f80/attachment.htm>
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Quoted 49 lines · Li Wang
Hi Liam,
Quoted 34 lines · Liam Howlett
Hello Li,
Thank you for looking at this testcase.
Quoted 13 lines · Li Wang
Hi Liam, Petr,
testcases/kernel/syscalls/brk/brk02.c line 36
Quoted 7 lines · Liam Howlett
...
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + }Quoted 6 lines
We got permission denied here while performing the brk02 on x86_64/s390x(kernel-4.18~). After looking at the manual page of
mprotect(), seems the access issue caused byPROT_EXEC."
POSIX says that the behavior ofmprotect()is unspecified if it isapplied
Quoted 7 lines
to a region of memory that was not obtained via mmap(2).
...
WhetherPROT_EXEChas any effect different fromPROT_READdepends on processor architecture, kernel version, and process state. IfREAD_IMPLIES_EXECis set in the process's personality flags (see personality(2)), specifyingPROT_READwill implicitly addPROT_EXEC. "Quoted 3 lines
Unforntunately, dropping the
PROT_EXECcauses the VMA behaviour to change and does not test what this testcase is intended to test.Yes, I agree with this. But am not sure if Linux take some action on security
side to prevent settingPROT_EXECpermission arbitrary.Quoted 1 line
Quoted 6 lines
Quoted 3 lines
Removing the
PROT_EXECand thePROT_WRITEdoes test what is supposed to be tested. Can you verify that this works on the s390x?Actually just removing the
PROT_EXECthen the brk02 can be passed on all my platforms.
Just removing the PROT_EXEC invalidates the test. However, if both PROT_EXEC and PROT_WRITE are removed, then the test still does what is intended.
Quoted 15 lines
Quoted 6 lines
Are you using real hardware to test this or can I use some sort of emulation to test on my side?
It looks like easily to reproduce.
I get failed with both virtual system and bare metal, e.g. the first one is on my Fedora33-workstation. But the worth to say, brk02 passed with raspberry pi3 and pi4.
x86_64
-------------
Quoted 8 lines
# virt-what # echo $? 0 # uname -r 5.10.22-200.fc33.x86_64 # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
This was my test platform. I also sent it to the Travis CI which passed for x86_64. I will re-examine the accepted code to ensure the cosmetic changes didn't invalidate my testing.
Quoted 23 lines
x86_64
-------------# virt-what kvm # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMAs390x
-------------# uname -r 5.12.0-rc4.vdso+ # virt-what ibm_systemz ibm_systemz-zvm # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMAarmv7l -- raspberry-pi3
-----------------------------
Quoted 13 lines
# uname -r 5.4.96-v7.1.el7 # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed armv7l -- raspberry-pi4 ----------------------------- # uname -rm 5.10.17-v7l+ armv7l # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed
Would you be willing to re-run the tests without both PROT_EXEC and PROT_WRITE?
Thank you,
Liam
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Liam,
Quoted 89 lines · Liam Howlett
Quoted 36 lines · Li Wang
Hi Liam,
Quoted 31 lines · Liam Howlett
Hello Li,
Thank you for looking at this testcase.
Quoted 13 lines · Li Wang
Hi Liam, Petr,
testcases/kernel/syscalls/brk/brk02.c line 36
Quoted 7 lines · Liam Howlett
...
testcases/kernel/syscalls/brk/brk02.c new file +5
-0+if (mprotect (addr - page_size, page_size, +PROT_READ |PROT_WRITE |PROT_EXEC )) { +tst_res (TFAIL ,"Cannot mprotect new VMA." ); +return ; + }Quoted 6 lines
We got permission denied here while performing the brk02 on x86_64/s390x(kernel-4.18~). After looking at the manual page of
mprotect(), seems the access issue caused byPROT_EXEC."
POSIX says that the behavior ofmprotect()is unspecified if it isapplied
Quoted 6 lines
to a region of memory that was not obtained via mmap(2).
...
WhetherPROT_EXEChas any effect different fromPROT_READdepends on processor architecture, kernel version, and process state. IfREAD_IMPLIES_EXECis set in the process's personality flags (see personality(2)), specifyingPROT_READwill implicitly add
PROT_EXEC.Quoted 13 lines
Quoted 6 lines
Quoted 1 line
"
Unforntunately, dropping the
PROT_EXECcauses the VMA behaviour to change and does not test what this testcase is intended to test.Yes, I agree with this. But am not sure if Linux take some action on security
side to prevent settingPROT_EXECpermission arbitrary.Quoted 1 line
Quoted 5 lines
Quoted 3 lines
Removing the
PROT_EXECand thePROT_WRITEdoes test what is supposed to be tested. Can you verify that this works on the s390x?Actually just removing the
PROT_EXECthen the brk02 can be passed on allmy
Quoted 1 line
platforms.
Just removing the
PROT_EXECinvalidates the test. However, if bothPROT_EXECandPROT_WRITEare removed, then the test still does what is intended.Quoted 15 lines
Quoted 6 lines
Are you using real hardware to test this or can I use some sort of emulation to test on my side?
It looks like easily to reproduce.
I get failed with both virtual system and bare metal, e.g. the first one is on my Fedora33-workstation. But the worth to say, brk02 passed with raspberry pi3 and pi4.
x86_64
-------------Quoted 8 lines
# virt-what # echo $? 0 # uname -r 5.10.22-200.fc33.x86_64 # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMA
Quoted 4 lines
This was my test platform. I also sent it to the Travis CI which passed for
x86_64. I will re-examine the accepted code to ensure the cosmetic changes didn't invalidate my testing.
FWIK, the Travis CI does not run test case, it just compiles/builds LTP across
many arches/platforms.
Quoted 39 lines
Quoted 25 lines
x86_64
-------------# virt-what kvm # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMAs390x
-------------# uname -r 5.12.0-rc4.vdso+ # virt-what ibm_systemz ibm_systemz-zvm # ./brk02 tst_test.c:1289: TINFO: Timeout per run is 0h 05m 00s brk02.c:41: TFAIL: Cannot mprotect new VMAarmv7l -- raspberry-pi3
-----------------------------Quoted 13 lines
# uname -r 5.4.96-v7.1.el7 # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed armv7l -- raspberry-pi4 ----------------------------- # uname -rm 5.10.17-v7l+ armv7l # ./brk02 tst_test.c:1291: TINFO: Timeout per run is 0h 05m 00s brk02.c:56: TPASS: munmap at least two VMAs of brk() passed
Quoted 3 lines
Would you be willing to re-run the tests without both
PROT_EXECandPROT_WRITE?
Yes, it will always PASS without 'PROT_EXEC|PROT_WRITE'.
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20210326/c86ad440/attachment-0001.htm>
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Liam, Li,
Quoted 7 lines
Quoted 3 lines
This was my test platform. I also sent it to the Travis CI which passed for
x86_64. I will re-examine the accepted code to ensure the cosmetic changes didn't invalidate my testing.FWIK, the Travis CI does not run test case, it just compiles/builds LTP across
many arches/platforms.
Yes, while KernelCI and some enterprise / embedded distros for their kernels run LTP testcases, the project does not run it. Cyril run some regression tests for few most important runtests before release manually. But having a server it'd be handy to run them.
Kind regards,
Petr
PATCH v2 1/1 brk02: Add test for removing more than one VMA
Quoted 6 lines · Petr Vorel
Hi Liam, Li,
Quoted 3 lines
Quoted 3 lines
This was my test platform. I also sent it to the Travis CI which passed for
x86_64. I will re-examine the accepted code to ensure the cosmetic changes didn't invalidate my testing.
Quoted 8 lines
Quoted 3 lines
FWIK, the Travis CI does not run test case, it just compiles/builds LTP across
many arches/platforms.Yes, while KernelCI and some enterprise / embedded distros for their kernels run LTP testcases, the project does not run it. Cyril run some regression tests for few most important runtests before release manually. But having a server it'd be handy to run them.
Thank you for clarification. What is the best way to re-test my changes? As I had said, I made a change that will not add EXEC but will still test removal of more than one VMA. We cannot just mprotect PROT_READ|PROT_WRITE as this will allow VMA merging to occur. My solution is to change the anon vma to just PROT_READ. It passes in my x86_64 test but since that was the case for me regardless, I cannot say that I have fixed the issue, but I have verified that the test still does what I expect it to do.
Thanks,
Liam
Patch below.
-------------------------------------------
+++ b/testcases/kernel/syscalls/brk/brk02.c
@@ -36,8 +36,7 @@ void brk_down_vmas(void)
return;
}- if (mprotect(addr - page_size, page_size,
- PROT_READ|PROT_WRITE|PROT_EXEC)) {
+ if (mprotect(addr - page_size, page_size, PROT_READ)) {
tst_res(TFAIL, "Cannot mprotect new VMA");
return;
}PATCH v2 1/1 brk02: Add test for removing more than one VMA
Hi Laim and all,
Quoted 16 lines · Liam Howlett
Quoted 3 lines · Petr Vorel
Hi Liam, Li,
Quoted 1 line
Quoted 1 line
This was my test platform. I also sent it to the Travis CI which
passed
Quoted 1 line
Quoted 1 line
Quoted 1 line
for
x86_64. I will re-examine the accepted code to ensure thecosmetic
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
changes didn't invalidate my testing.
FWIK, the Travis CI does not run test case, it just compiles/builds LTP across
many arches/platforms.Yes, while KernelCI and some enterprise / embedded distros for their
kernels run
Quoted 1 line
LTP testcases, the project does not run it. Cyril run some regression
Quoted 9 lines
tests for
Quoted 1 line
few most important runtests before release manually. But having a server
it'd be
Quoted 1 line
handy to run them.
Thank you for clarification. What is the best way to re-test my changes? As I had said, I made a change that will not add EXEC but will
No best way I guess, it would be great if you can do more arches verification before patch sending, but then the maintainers would help do that also.
Quoted 7 lines
still test removal of more than one VMA. We cannot just mprotect PROT_READ|PROT_WRITE as this will allow VMA merging to occur. My solution is to change the anon vma to just
PROT_READ. It passes in myx86_64test but since that was the case for me regardless, I cannot say that I have fixed the issue, but I have verified that the test still does what I expect it to do.
I'm fine with just test PROT_READ. And we can take this way to fix the FAIL before the next LTP release.
But there is still a query in my mind, whether the FAIL I mentioned before is a kernel bug or just caused by preventing process Self Modiefed Code, that probably needs more investigation.
Quoted 21 lines
Thanks,
LiamPatch below.
-------------------------------------------+++
b/testcases/kernel/syscalls/brk/brk02.c(quoted diff) +1
-2@@ -36,8 +36,7 @@ void brk_down_vmas(void) return; } - if (mprotect(addr - page_size, page_size, - PROT_READ|PROT_WRITE|PROT_EXEC)) { + if (mprotect(addr - page_size, page_size, PROT_READ)) { tst_res(TFAIL, "Cannot mprotect new VMA"); return; }
Signature · 5 lines
Regards, Li Wang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linux.it/pipermail/ltp/attachments/20210507/5b376d68/attachment.htm>