Thank you for your contribution! Sashiko AI review found 10 potential issues to consider: - [Medium] selftests/dmabuf-heaps: pidfd_alloc_supported() mishandles disabled memcg - [Medium] selftests/cgroup: dmabuf tests read memory.stat below rstat flush threshold - [Medium] selftests/dmabuf-heaps: open_self_cgroupfd() misses memory controller check - [Medium] selftests/dmabuf-heaps: check_deps() runs before handle_build() in vmtest.sh - [Medium] selftests/cgroup: test_memcg_dmabuf_xfer_at_limit() fails instead of skipping - [Low] selftests/cgroup: child_cg leak on cg_create() error in test_memcg_dmabuf() - [Low] selftests/cgroup: missing skip checks in test_memcg_dmabuf_xfer_at_limit() - [Low] selftests/cgroup: unused MEMCG_SOCKSTAT_WAIT_RETRIES macro in test_memcontrol - [Low] selftests/dmabuf-heaps: vmtest.sh creates TMP_DIR before setting EXIT trap - [Low] selftests/dmabuf-heaps: make_runner() omits test_memcg_dmabuf_dead_cgroup --
commit 9485d75462eea316c6a545e3ba3f2d0923e12b75 Author: Albert Esteve aesteve@redhat.com
selftests/dmabuf-heaps: Add dma-buf memcg accounting tests
This commit adds selftests for dma-buf memcg accounting using charge_fd and DMA_BUF_IOCTL_XFER_CHARGE in dmabuf-heaps and cgroup/test_memcontrol, along with a virtme-ng test runner script.
diff --git a/tools/testing/selftests/cgroup/test_memcontrol.c b/tools/testing/selftests/cgroup/test_memcontrol.c index bb673012dcf08..f612726bd9af7 100644 --- a/tools/testing/selftests/cgroup/test_memcontrol.c +++ b/tools/testing/selftests/cgroup/test_memcontrol.c @@ -27,6 +34,13 @@ #define MEMCG_FLUSH_BYTES \ (2 * sysconf(_SC_PAGESIZE) * MEMCG_CHARGE_BATCH * get_nprocs()) +#define DMA_HEAP_SYSTEM "/dev/dma_heap/system" +#define ONE_MEG (1024 * 1024)
+#define MEMCG_SOCKSTAT_WAIT_RETRIES 30
[Severity: Low] MEMCG_SOCKSTAT_WAIT_RETRIES was removed in eb69e955de7b ("selftests: cgroup: make test_memcg_sock check flushed sock stats") and has no users here. Was this re-added by accident during a rebase?
+#define MEMCG_DMABUF_WAIT_RETRIES 30 +#define MEMCG_DMABUF_DEAD_WAIT_RETRIES 80
static bool has_localevents; static bool has_recursiveprot; static int page_size; @@ -1797,6 +1811,417 @@ static int test_memcg_inotify_delete_dir(const char *root) return ret; } +static int memcg_dmabuf_child(const char *cgroup, void *arg) +{
- pid_t *parent_pid = arg;
- prctl(PR_SET_PDEATHSIG, SIGKILL);
- if (getppid() != *parent_pid)
return 0;- pause();
- return 0;
+}
+/*
- This test allocates a dma-buf via DMA_HEAP_IOCTL_ALLOC with a pidfd
- pointing to a child process in a separate cgroup, then checks that
- memory.stat[dmabuf] in the child's cgroup rises by the allocation size
- and returns to zero after the buffer fd is closed.
- */
+static int test_memcg_dmabuf(const char *root) +{
- char *parent = NULL, *child_cg = NULL;
- int ret = KSFT_FAIL;
- int heap_fd = -1, dmabuf_fd = -1, pidfd = -1;
- pid_t child_pid, parent_pid = getpid();
- int child_status;
- long dmabuf_stat;
- struct dma_heap_allocation_data alloc = {
.len = ONE_MEG,.fd_flags = O_RDWR | O_CLOEXEC,- };
- if (access(DMA_HEAP_SYSTEM, R_OK | W_OK)) {
ret = KSFT_SKIP;goto cleanup;- }
- parent = cg_name(root, "dmabuf_memcg_test");
- if (!parent)
goto cleanup;- if (cg_create(parent))
goto cleanup_parent;- if (cg_write(parent, "cgroup.subtree_control", "+memory"))
goto cleanup_parent;- child_cg = cg_name(parent, "child");
- if (!child_cg)
goto cleanup_parent;- if (cg_create(child_cg))
goto cleanup_parent;
[Severity: Low] Does this leak child_cg when cg_create(child_cg) fails? Jumping to cleanup_parent skips free(child_cg) in cleanup_child.
[ ... ]
- if (ioctl(heap_fd, DMA_HEAP_IOCTL_ALLOC, &alloc) < 0) {
close(alloc.charge_fd);goto cleanup_heap;- }
[Severity: Medium] Should test_memcg_dmabuf_xfer_at_limit() set ret = KSFT_SKIP when DMA_HEAP_IOCTL_ALLOC fails with EINVAL, ENOTTY or EOPNOTSUPP, matching test_memcg_dmabuf() and test_memcg_dmabuf_dead_cgroup()? Without that, the test fails instead of skipping on kernels without charge_fd support.
- close(alloc.charge_fd);
- dmabuf_fd = (int)alloc.fd;
- stat_a = cg_read_key_long(cg_a, "memory.stat", "dmabuf ");
- if (stat_a == -1) {
ret = KSFT_SKIP;goto cleanup_dmabuf;- }
- if (stat_a != ONE_MEG)
stat_a = cg_read_key_long_poll(cg_a, "memory.stat","dmabuf ", ONE_MEG,MEMCG_DMABUF_WAIT_RETRIES,DEFAULT_WAIT_INTERVAL_US);- if (stat_a != ONE_MEG) {
fprintf(stderr, "Charge not in A after alloc: expected %d, got %ld\n",ONE_MEG, stat_a);goto cleanup_dmabuf;- }
- cg_b_fd = open(cg_b, O_RDONLY | O_DIRECTORY);
- if (cg_b_fd < 0)
goto cleanup_dmabuf;- xfer.charge_fd = (__u32)cg_b_fd;
- if (ioctl(dmabuf_fd, DMA_BUF_IOCTL_XFER_CHARGE, &xfer) == 0) {
fprintf(stderr, "xfer_charge should have failed at limit\n");goto cleanup_cgfd;- }
- if (errno != ENOMEM) {
fprintf(stderr, "Expected ENOMEM, got %s\n", strerror(errno));goto cleanup_cgfd;- }
[Severity: Low] Can DMA_BUF_IOCTL_XFER_CHARGE also fail with ENOTTY or EOPNOTSUPP when the ioctl or memcg accounting is not supported, and should those also set ret = KSFT_SKIP here?
- stat_a = cg_read_key_long_poll(cg_a, "memory.stat",
"dmabuf ", ONE_MEG,MEMCG_DMABUF_WAIT_RETRIES,DEFAULT_WAIT_INTERVAL_US);- if (stat_a != ONE_MEG) {
fprintf(stderr, "Charge leaked from A: expected %d, got %ld\n",ONE_MEG, stat_a);goto cleanup_cgfd;- }
[Severity: Medium] Does polling memory.stat for ONE_MEG leave rstat unflushed on machines with 4 or more CPUs?
Reading memory.stat calls __mem_cgroup_flush_stats(memcg, false), which only flushes rstat when pending updates exceed MEMCG_CHARGE_BATCH * num_online_cpus() (64 pages per CPU). ONE_MEG is 256 pages on 4 KB page systems, so on >= 4 CPUs a 1 MB charge or uncharge in test_memcg_dmabuf(), test_memcg_dmabuf_xfer_at_limit() and test_memcg_dmabuf_dead_cgroup() does not trigger a synchronous flush and instead waits up to 3 seconds for the deferrable stats_flush_dwork worker, as described in eb69e955de7b ("selftests: cgroup: make test_memcg_sock check flushed sock stats").
In addition, because stat_a already read ONE_MEG before DMA_BUF_IOCTL_XFER_CHARGE, cg_read_key_long_poll() here sees the cached ONE_MEG value on its first iteration and returns immediately without ensuring any rstat flush happened after the failed transfer. Using MEMCG_FLUSH_BYTES for the buffer size (or forcing a flush before reading) would make the memory.stat reads cross the flush threshold.
diff --git a/tools/testing/selftests/dmabuf-heaps/dmabuf-heap.c b/tools/testing/selftests/dmabuf-heaps/dmabuf-heap.c index 1d49df6719198..44e3f33acf877 100644 --- a/tools/testing/selftests/dmabuf-heaps/dmabuf-heap.c +++ b/tools/testing/selftests/dmabuf-heaps/dmabuf-heap.c
[ ... ]
+/*
- Probe whether the kernel honours charge_fd in DMA_HEAP_IOCTL_ALLOC.
- */
+static bool pidfd_alloc_supported(int heap_fd) +{
- int devnull_fd, dmabuf_fd = -1, ret;
- devnull_fd = open("/dev/null", O_RDONLY);
- if (devnull_fd < 0)
return false;- ret = dmabuf_heap_alloc_pidfd(heap_fd, ONE_MEG, 0, devnull_fd, &dmabuf_fd);
- if (dmabuf_fd >= 0) {
close(dmabuf_fd);dmabuf_fd = -1;- }
- close(devnull_fd);
- return ret < 0;
+}
[Severity: Medium] Does pidfd_alloc_supported() return true when memcg is disabled?
When CONFIG_MEMCG=n or cgroup_disable=memory is passed at boot, get_obj_cgroup_from_fd() returns ERR_PTR(-EOPNOTSUPP) before inspecting the fd, whereas an active memcg setup rejects /dev/null with EBADF. Because ret < 0 is true on EOPNOTSUPP, pidfd_alloc_supported() reports that charge_fd is supported and the subsequent allocation tests run and fail instead of skipping. Checking ret < 0 && errno == EBADF avoids this.
[ ... ]
+static int open_self_cgroupfd(void) +{
- char cgroup_path[PATH_MAX];
- char line[PATH_MAX];
- int ret = -1;
- FILE *f;
- f = fopen("/proc/self/cgroup", "r");
- if (!f)
return -1;- while (fgets(line, sizeof(line), f)) {
if (strncmp(line, "0::", 3))continue;line[strcspn(line, "\n")] = '\0';if (snprintf(cgroup_path, sizeof(cgroup_path),"/sys/fs/cgroup%s", line + 3) < PATH_MAX)ret = 0;break;- }
- fclose(f);
- if (ret)
return -1;- return open(cgroup_path, O_RDONLY | O_DIRECTORY);
+}
[Severity: Medium] Can test_alloc_cgroupfd_self() and test_xfer_charge_cgroupfd_self() fail with ENOENT when the test process runs in a cgroup without the memory controller enabled?
Unlike the pidfd path in get_obj_cgroup_from_fd(), which walks up to the effective ancestor via task_css(), the cgroupfd path calls css_tryget_online_from_dir(dentry, &memory_cgrp_subsys) directly on the opened directory and returns -ENOENT when +memory is not enabled in its parent's cgroup.subtree_control. Verifying that memory.current (or memory.stat) exists in cgroup_path, or skipping on ENOENT, would prevent false failures when dmabuf-heap is run outside vmtest.sh.
diff --git a/tools/testing/selftests/dmabuf-heaps/vmtest.sh b/tools/testing/selftests/dmabuf-heaps/vmtest.sh new file mode 100755 index 0000000000000..3574b71ca83e8 --- /dev/null +++ b/tools/testing/selftests/dmabuf-heaps/vmtest.sh @@ -0,0 +1,226 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0 +# +# Copyright (c) 2026 Red Hat +# +# Dependencies: +# * virtme-ng +# * qemu (used by virtme-ng)
+readonly SCRIPT_DIR="$(cd -P -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd -P)" +readonly KERNEL_CHECKOUT=$(realpath "${SCRIPT_DIR}"/../../../../) +readonly CGROUP_DIR="${KERNEL_CHECKOUT}/tools/testing/selftests/cgroup"
+source "${SCRIPT_DIR}"/../kselftest/ktap_helpers.sh
+readonly DMABUF_HEAP_TEST="${SCRIPT_DIR}"/dmabuf-heap +readonly MEMCONTROL_TEST="${CGROUP_DIR}"/test_memcontrol +readonly TMP_DIR=$(mktemp -d /tmp/dmabuf-vmtest.XXXXXXXX)
[Severity: Low] TMP_DIR is created before getopts and before `trap cleanup EXIT` is registered, so running `vmtest.sh -h` or passing an invalid option exits in usage() and leaves the temporary directory behind in /tmp.
[ ... ]
+check_deps() {
- for dep in vng make; do
if [[ ! -x $(command -v "${dep}") ]]; thenecho -e "skip: dependency ${dep} not found!\n"exit "${KSFT_SKIP}"fi- done
- if ${SHELL_MODE}; then
return- fi
- if [[ ! -x "${DMABUF_HEAP_TEST}" ]]; then
printf "skip: %s not found!" "${DMABUF_HEAP_TEST}"printf " Please build the kselftest dmabuf-heaps target (or use -b).\n"exit "${KSFT_SKIP}"- fi
- if [[ ! -x "${MEMCONTROL_TEST}" ]]; then
printf "skip: %s not found!" "${MEMCONTROL_TEST}"printf " Please build the kselftest cgroup target (or use -b).\n"exit "${KSFT_SKIP}"- fi
+}
[Severity: Medium] Does `vmtest.sh -b` exit in check_deps() before building the tests when DMABUF_HEAP_TEST or MEMCONTROL_TEST has not been built yet? check_deps() is called before handle_build(), and only SHELL_MODE returns early before checking that the test binaries already exist.
[ ... ]
- run "dmabuf-heap charge_fd ioctl" ${DMABUF_HEAP_TEST}
- run "memcontrol dma-buf memcg" ${MEMCONTROL_TEST} test_memcg_dmabuf
- run "memcontrol dma-buf xfer limit" \
${MEMCONTROL_TEST} test_memcg_dmabuf_xfer_at_limit
[Severity: Low] Should make_runner() also run test_memcg_dmabuf_dead_cgroup here?