All of lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Albert Esteve" <aesteve@redhat.com>, "Tejun Heo" <tj@kernel.org>,
	"Johannes Weiner" <hannes@cmpxchg.org>,
	"Michal Koutný" <mkoutny@suse.com>,
	"Jonathan Corbet" <corbet@lwn.net>,
	"Shuah Khan" <skhan@linuxfoundation.org>,
	"Sumit Semwal" <sumit.semwal@linaro.org>,
	"Christian König" <christian.koenig@amd.com>,
	"Michal Hocko" <mhocko@kernel.org>,
	"Roman Gushchin" <roman.gushchin@linux.dev>,
	"Shakeel Butt" <shakeel.butt@linux.dev>,
	"Muchun Song" <muchun.song@linux.dev>,
	"Andrew Morton" <akpm@linux-foundation.org>,
	"Benjamin Gaignard" <benjamin.gaignard@collabora.com>,
	"Brian Starkey" <Brian.Starkey@arm.com>,
	"John Stultz" <jstultz@google.com>,
	"T.J. Mercier" <tjmercier@google.com>,
	"Christian Brauner" <brauner@kernel.org>,
	"Paul Moore" <paul@paul-moore.com>,
	"James Morris" <jmorris@namei.org>,
	"Serge E. Hallyn" <serge@hallyn.com>,
	"Stephen Smalley" <stephen.smalley.work@gmail.com>,
	"Shuah Khan" <shuah@kernel.org>,
	"Randy Dunlap" <rdunlap@infradead.org>,
	"Ondrej Mosnáček" <omosnacek@gmail.com>,
	"Jens Wiklander" <jenswi@kernel.org>,
	"Sumit Garg" <sumit.garg@kernel.org>,
	"Philipp Stanner" <phasta@kernel.org>
Cc: airlied@gmail.com, dri-devel@lists.freedesktop.org,
	linux-mm@kvack.org, linux-security-module@vger.kernel.org,
	selinux@vger.kernel.org, op-tee@lists.trustedfirmware.org,
	echanude@redhat.com, mripard@kernel.org, media-ci@linuxtv.org
Subject: Re: [PATCH v5 8/8] selftests/dmabuf-heaps: Add dma-buf memcg accounting tests
Date: Wed, 18 Nov 2026 08:41:04 +0000	[thread overview]
Message-ID: <sashiko-outbox-220474@kernel.org> (raw)
In-Reply-To: <20261117-v2_20230123_tjmercier_google_com-v5-8-1ce975f2a9ad@redhat.com>

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}") ]]; then
> +			echo -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?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117-v2_20230123_tjmercier_google_com-v5-0-1ce975f2a9ad@redhat.com?part=8

WARNING: multiple messages have this Message-ID (diff)
From: sashiko-bot--- via OP-TEE <op-tee@lists.trustedfirmware.org>
To: "Albert Esteve" <aesteve@redhat.com>, "Tejun Heo" <tj@kernel.org>,
	"Johannes Weiner" <hannes@cmpxchg.org>,
	"Michal Koutný" <mkoutny@suse.com>,
	"Jonathan Corbet" <corbet@lwn.net>,
	"Shuah Khan" <skhan@linuxfoundation.org>,
	"Sumit Semwal" <sumit.semwal@linaro.org>,
	"Christian König" <christian.koenig@amd.com>,
	"Michal Hocko" <mhocko@kernel.org>,
	"Roman Gushchin" <roman.gushchin@linux.dev>,
	"Shakeel Butt" <shakeel.butt@linux.dev>,
	"Muchun Song" <muchun.song@linux.dev>,
	"Andrew Morton" <akpm@linux-foundation.org>,
	"Benjamin Gaignard" <benjamin.gaignard@collabora.com>,
	"Brian Starkey" <Brian.Starkey@arm.com>,
	"John Stultz" <jstultz@google.com>,
	"T.J. Mercier" <tjmercier@google.com>,
	"Christian Brauner" <brauner@kernel.org>,
	"Paul Moore" <paul@paul-moore.com>,
	"James Morris" <jmorris@namei.org>,
	"Serge E. Hallyn" <serge@hallyn.com>,
	"Stephen Smalley" <stephen.smalley.work@gmail.com>,
	"Shuah Khan" <shuah@kernel.org>,
	"Randy Dunlap" <rdunlap@infradead.org>,
	"Ondrej Mosnáček" <omosnacek@gmail.com>,
	"Jens Wiklander" <jenswi@kernel.org>,
	"Sumit Garg" <sumit.garg@kernel.org>,
	"Philipp Stanner" <phasta@kernel.org>
Cc: airlied@gmail.com, dri-devel@lists.freedesktop.org,
	linux-mm@kvack.org, linux-security-module@vger.kernel.org,
	selinux@vger.kernel.org, op-tee@lists.trustedfirmware.org,
	echanude@redhat.com, mripard@kernel.org, media-ci@linuxtv.org
Subject: Re: [PATCH v5 8/8] selftests/dmabuf-heaps: Add dma-buf memcg accounting tests
Date: Wed, 18 Nov 2026 08:41:04 +0000	[thread overview]
Message-ID: <sashiko-outbox-220474@kernel.org> (raw)
In-Reply-To: <20261117-v2_20230123_tjmercier_google_com-v5-8-1ce975f2a9ad@redhat.com>

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}") ]]; then
> +			echo -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?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117-v2_20230123_tjmercier_google_com-v5-0-1ce975f2a9ad@redhat.com?part=8

  reply	other threads:[~2026-11-18  8:41 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17  8:39 [PATCH v5 0/8] memcg: dma-buf cgroup accounting and charge transfer Albert Esteve
2026-11-17  8:39 ` Albert Esteve via OP-TEE
2026-11-17  8:39 ` [PATCH v5 1/8] memcg: Add MEMCG_DMABUF accounting infrastructure Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:40   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 2/8] dma-heap: charge dma-buf memory via explicit memcg Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:40   ` sashiko-bot
2026-11-18  8:40     ` sashiko-bot--- via OP-TEE
2026-11-17  8:39 ` [PATCH v5 3/8] dma-buf: transfer objcg charges via DMA_BUF_IOCTL_XFER_CHARGE Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:41   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 4/8] lsm: Add memcg_charge_task LSM hook Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-17 22:54   ` Paul Moore
2026-11-18  8:41   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 5/8] lsm: Add memcg_charge_cgroup " Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:41   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 6/8] selinux: Restrict cross-cgroup memcg charging Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:41   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 7/8] selinux: Restrict memcg charging to a cgroupfd target Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:41   ` sashiko-bot
2026-11-17  8:39 ` [PATCH v5 8/8] selftests/dmabuf-heaps: Add dma-buf memcg accounting tests Albert Esteve
2026-11-17  8:39   ` Albert Esteve via OP-TEE
2026-11-18  8:41   ` sashiko-bot [this message]
2026-11-18  8:41     ` sashiko-bot--- via OP-TEE
2026-11-17 22:54 ` [PATCH v5 0/8] memcg: dma-buf cgroup accounting and charge transfer Paul Moore

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-220474@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=Brian.Starkey@arm.com \
    --cc=aesteve@redhat.com \
    --cc=airlied@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=benjamin.gaignard@collabora.com \
    --cc=brauner@kernel.org \
    --cc=christian.koenig@amd.com \
    --cc=corbet@lwn.net \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=echanude@redhat.com \
    --cc=hannes@cmpxchg.org \
    --cc=jenswi@kernel.org \
    --cc=jmorris@namei.org \
    --cc=jstultz@google.com \
    --cc=linux-mm@kvack.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=media-ci@linuxtv.org \
    --cc=mhocko@kernel.org \
    --cc=mkoutny@suse.com \
    --cc=mripard@kernel.org \
    --cc=muchun.song@linux.dev \
    --cc=omosnacek@gmail.com \
    --cc=op-tee@lists.trustedfirmware.org \
    --cc=paul@paul-moore.com \
    --cc=phasta@kernel.org \
    --cc=rdunlap@infradead.org \
    --cc=roman.gushchin@linux.dev \
    --cc=sashiko-reviews@lists.linux.dev \
    --cc=selinux@vger.kernel.org \
    --cc=serge@hallyn.com \
    --cc=shakeel.butt@linux.dev \
    --cc=shuah@kernel.org \
    --cc=skhan@linuxfoundation.org \
    --cc=stephen.smalley.work@gmail.com \
    --cc=sumit.garg@kernel.org \
    --cc=sumit.semwal@linaro.org \
    --cc=tj@kernel.org \
    --cc=tjmercier@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.