tests/tcg/multiarch: add vma-pthread.c
A message that is not in the archive (sent off-list or not yet received): 20221223115348.tgfwdlektsulebxk@heavy
PATCH 1/1 tests/tcg/multiarch: add vma-pthread.c
Add a test that locklessly changes and exercises page protection bits from various threads. This helps catch race conditions in the VMA handling.
--- tests/tcg/multiarch/Makefile.target | 3 + tests/tcg/multiarch/munmap-pthread.c | 16 +-- tests/tcg/multiarch/nop_func.h | 25 ++++ tests/tcg/multiarch/vma-pthread.c | 185 +++++++++++++++++++++++++++ 4 files changed, 214 insertions(+), 15 deletions(-) create mode 100644 tests/tcg/multiarch/nop_func.h create mode 100644 tests/tcg/multiarch/vma-pthread.c
tests/tcg/multiarch/Makefile.target +3 -0
@@ -39,6 +39,9 @@ signals: LDFLAGS+=-lrt -lpthread munmap-pthread: CFLAGS+=-pthread munmap-pthread: LDFLAGS+=-pthread +vma-pthread: CFLAGS+=-pthread +vma-pthread: LDFLAGS+=-pthread + # We define the runner for test-mmap after the individual # architectures have defined their supported pages sizes. If no # additional page sizes are defined we only run the default test.
tests/tcg/multiarch/munmap-pthread.c +1 -15
@@ -7,21 +7,7 @@#include <sys/mman.h> #include <unistd.h> -static const char nop_func[] = { -#if defined(__aarch64__) -0xc0 ,0x03 ,0x5f ,0xd6 ,/* ret */ -#elif defined(__alpha__) -0x01 ,0x80 ,0xFA ,0x6B ,/* ret */ -#elif defined(__arm__) -0x1e ,0xff ,0x2f ,0xe1 ,/* bx lr */ -#elif defined(__riscv) -0x67 ,0x80 ,0x00 ,0x00 ,/* ret */ -#elif defined(__s390__) -0x07 ,0xfe ,/* br %r14 */ -#elif defined(__i386__) || defined(__x86_64__) -0xc3 ,/* ret */ -#endif -}; +#include "nop_func.h" static void *thread_mmap_munmap (void *arg) {
tests/tcg/multiarch/nop_func.h new file +25 -0
@@ -0,0 +1,25 @@ +/* +* No-op functions that can be safely copied. +* +* SPDX-License-Identifier: GPL-2.0-or-later +*/ +#ifndef NOP_FUNC_H +#define NOP_FUNC_H + +static const char nop_func[] = { +#if defined(__aarch64__) +0xc0 ,0x03 ,0x5f ,0xd6 ,/* ret */ +#elif defined(__alpha__) +0x01 ,0x80 ,0xFA ,0x6B ,/* ret */ +#elif defined(__arm__) +0x1e ,0xff ,0x2f ,0xe1 ,/* bx lr */ +#elif defined(__riscv) +0x67 ,0x80 ,0x00 ,0x00 ,/* ret */ +#elif defined(__s390__) +0x07 ,0xfe ,/* br %r14 */ +#elif defined(__i386__) || defined(__x86_64__) + 0xc3, /* ret */ +#endif +}; + +#endif
tests/tcg/multiarch/vma-pthread.c new file +185 -0
185 lines not on this page · raw message
Signature · 1 line
2.38.1
Quoted 11 lines · Ilya Leoshkevich
Add a test that locklessly changes and exercises page protection bits from various threads. This helps catch race conditions in the VMA handling.
Signed-off-by Ilya Leoshkevich <[email protected]>---
tests/tcg/multiarch/Makefile.target | 3 +
tests/tcg/multiarch/munmap-pthread.c | 16 +--
tests/tcg/multiarch/nop_func.h | 25 ++++
tests/tcg/multiarch/vma-pthread.c | 185
+++++++++++++++++++++++++++
Quoted 3 lines
4 files changed, 214 insertions(+), 15 deletions(-)create mode 100644
tests/tcg/multiarch/nop_func.hcreate mode 100644tests/tcg/multiarch/vma-pthread.c
This was meant to be a reply to the bug report for [1], but apparently I forgot to Cc: the mailing list. Copying the original message here:
---
Hi,
Wasmtime testsuite started failing randomly, complaining that clock_gettime() returns -EFAULT. Bisect points to this commit.
I could not see anything obviously wrong here with the manual review, and the failure was not reproducible when running individual testcases or using strace. So I wrote a stress test (which I will post shortly), which runs fine on the host, but reproduces the issue with qemu-user.
When run with -strace, it also triggers an assertion:
qemu-x86_64: ../accel/tcg/tb-maint.c:595:
tb_invalidate_phys_page_unwind: Assertion `pc != 0' failed.
qemu-x86_64: /home/iii/qemu/include/qemu/rcu.h:102:rcu_read_unlock: Assertion `p_rcu_reader->depth != 0' failed.
I haven't tried analyzing what is causing all this yet, but at least now the reproducer is small (~200LOC) and fails faster than 1s.
Best regards,
Ilya
---
[1] https://lists.gnu.org/archive/html/qemu-devel/2022-12/msg03615.html
Quoted 6 lines · Ilya Leoshkevich
tests/tcg/multiarch/vma-pthread.c new file +6
-0+static void *thread_write (void *arg) +{ +struct context *ctx = arg; +struct timespec *ts; +size_t i, j; +int ret;
tests/tcg/multiarch/vma-pthread.c line 77
Quoted 10 lines
tests/tcg/multiarch/vma-pthread.c new file +10
-0+ +for (i =0 ; ctx->mutator_count; i++) { + j = (i &PAGE_IDX_MASK ) |PAGE_IDX_W_MASK ; +/* Write directly. */ +memcpy (&ctx->ptr[j * ctx->pagesize], nop_func,sizeof (nop_func)); +/* Write using a syscall. */ + ts = (struct timespec *)(&ctx->ptr[(j +1 ) * ctx->pagesize] - +sizeof (struct timespec )); + ret =clock_gettime (CLOCK_REALTIME , ts); +assert (ret ==0 );
With the 3 issues that you pointed out in the other email, this is the only remaining failure. This happens because of two issues:
(1) When checking for writability, we actually check for both read+write:
#define VERIFY_WRITE (PAGE_READ | PAGE_WRITE)
This is very likely a bug, but we'd need to audit all uses to find out where we might really want read+write.
Quoted 18 lines
tests/tcg/multiarch/vma-pthread.c new file +18
-0+static void *thread_mutate (void *arg) +{ +size_t i, start_idx, end_idx, page_idx, tmp; +struct context *ctx = arg; +unsigned int seed; +int prot, ret; + + seed = (unsigned int )time (NULL ); +for (i =0 ; i <50000 ; i++) { + start_idx =rand_r (&seed) &PAGE_IDX_MASK ; + end_idx =rand_r (&seed) &PAGE_IDX_MASK ; +if (start_idx > end_idx) { + tmp = start_idx; + start_idx = end_idx; + end_idx = tmp; + } + prot =rand_r (&seed) & (PROT_READ |PROT_WRITE |PROT_EXEC ); + for (page_idx = start_idx & REGION_MASK; page_idx <= end_idx;
tests/tcg/multiarch/vma-pthread.c line 123
Quoted 10 lines
tests/tcg/multiarch/vma-pthread.c new file +10
-0+ page_idx += PAGES_PER_REGION) { +if (page_idx &PAGE_IDX_R_MASK ) { + prot |=PROT_READ ; + } +if (page_idx &PAGE_IDX_W_MASK ) { + prot |=PROT_WRITE ; + } +if (page_idx &PAGE_IDX_X_MASK ) { + prot |=PROT_EXEC ; + }
... and here we can wind up with write-only pages.
(2) Certain hardware, like x86_64, does not support write-only pages -- writable implies readable -- so the testcase runs on some hardware. It is a bug that we don't model this as well.
r~