- Linux-stable-mirror - lists.linaro.org

[PATCH] hwmon: (dell-smm): Fix off-by-one error in dell_smm_is_visible()

by Armin Wolf

The documentation states that on machines supporting only global fan mode control, the pwmX_enable attributes should only be created for the first fan channel (pwm1_enable, aka channel 0). Fix the off-by-one error caused by the fact that fan channels have a zero-based index. Cc: stable(a)vger.kernel.org Fixes: 1c1658058c99 ("hwmon: (dell-smm) Add support for automatic fan mode") Signed-off-by: Armin Wolf <W_Armin(a)gmx.de> --- drivers/hwmon/dell-smm-hwmon.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/hwmon/dell-smm-hwmon.c b/drivers/hwmon/dell-smm-hwmon.c index 683baf361c4c..a34753fc2973 100644 --- a/drivers/hwmon/dell-smm-hwmon.c +++ b/drivers/hwmon/dell-smm-hwmon.c @@ -861,9 +861,9 @@ static umode_t dell_smm_is_visible(const void *drvdata, enum hwmon_sensor_types if (auto_fan) { /* * The setting affects all fans, so only create a - * single attribute. + * single attribute for the first fan channel. */ - if (channel != 1) + if (channel != 0) return 0; /* -- 2.39.5

1 week, 1 day

2
1
0 0

[PATCH] hwmon: (w83791d) Convert macros to functions to avoid TOCTOU

by Gui-Dong Han

The macro FAN_FROM_REG evaluates its arguments multiple times. When used in lockless contexts involving shared driver data, this leads to Time-of-Check to Time-of-Use (TOCTOU) race conditions, potentially causing divide-by-zero errors. Convert the macro to a static function. This guarantees that arguments are evaluated only once (pass-by-value), preventing the race conditions. Additionally, in store_fan_div, move the calculation of the minimum limit inside the update lock. This ensures that the read-modify-write sequence operates on consistent data. Adhere to the principle of minimal changes by only converting macros that evaluate arguments multiple times and are used in lockless contexts. Link: https://lore.kernel.org/all/CALbr=LYJ_ehtp53HXEVkSpYoub+XYSTU8Rg=o1xxMJ8=5z… Fixes: 9873964d6eb2 ("[PATCH] HWMON: w83791d: New hardware monitoring driver for the Winbond W83791D") Cc: stable(a)vger.kernel.org Signed-off-by: Gui-Dong Han <hanguidong02(a)gmail.com> --- Based on the discussion in the link, I will submit a series of patches to address TOCTOU issues in the hwmon subsystem by converting macros to functions or adjusting locking where appropriate. --- drivers/hwmon/w83791d.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) diff --git a/drivers/hwmon/w83791d.c b/drivers/hwmon/w83791d.c index ace854b370a0..996e36951f9d 100644 --- a/drivers/hwmon/w83791d.c +++ b/drivers/hwmon/w83791d.c @@ -218,9 +218,14 @@ static u8 fan_to_reg(long rpm, int div) return clamp_val((1350000 + rpm * div / 2) / (rpm * div), 1, 254); } -#define FAN_FROM_REG(val, div) ((val) == 0 ? -1 : \ - ((val) == 255 ? 0 : \ - 1350000 / ((val) * (div)))) +static int fan_from_reg(int val, int div) +{ + if (val == 0) + return -1; + if (val == 255) + return 0; + return 1350000 / (val * div); +} /* for temp1 which is 8-bit resolution, LSB = 1 degree Celsius */ #define TEMP1_FROM_REG(val) ((val) * 1000) @@ -521,7 +526,7 @@ static ssize_t show_##reg(struct device *dev, struct device_attribute *attr, \ struct w83791d_data *data = w83791d_update_device(dev); \ int nr = sensor_attr->index; \ return sprintf(buf, "%d\n", \ - FAN_FROM_REG(data->reg[nr], DIV_FROM_REG(data->fan_div[nr]))); \ + fan_from_reg(data->reg[nr], DIV_FROM_REG(data->fan_div[nr]))); \ } show_fan_reg(fan); @@ -585,10 +590,10 @@ static ssize_t store_fan_div(struct device *dev, struct device_attribute *attr, if (err) return err; + mutex_lock(&data->update_lock); /* Save fan_min */ - min = FAN_FROM_REG(data->fan_min[nr], DIV_FROM_REG(data->fan_div[nr])); + min = fan_from_reg(data->fan_min[nr], DIV_FROM_REG(data->fan_div[nr])); - mutex_lock(&data->update_lock); data->fan_div[nr] = div_to_reg(nr, val); switch (nr) { -- 2.43.0

1 week, 1 day

2
1
0 0

[PATCH AUTOSEL 6.18-5.10] ACPI: property: Use ACPI functions in acpi_graph_get_next_endpoint() only

by Sasha Levin

From: Sakari Ailus <sakari.ailus(a)linux.intel.com> [ Upstream commit 5d010473cdeaabf6a2d3a9e2aed2186c1b73c213 ] Calling fwnode_get_next_child_node() in ACPI implementation of the fwnode property API is somewhat problematic as the latter is used in the impelementation of the former. Instead of using fwnode_get_next_child_node() in acpi_graph_get_next_endpoint(), call acpi_get_next_subnode() directly instead. Signed-off-by: Sakari Ailus <sakari.ailus(a)linux.intel.com> Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas(a)ideasonboard.com> Reviewed-by: Jonathan Cameron <jonathan.cameron(a)huawei.com> Link: https://patch.msgid.link/20251001104320.1272752-3-sakari.ailus@linux.intel.… Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- LLM Generated explanations, may be completely bogus: ## Analysis ### 1. COMMIT MESSAGE ANALYSIS The commit message states: - Problem: `acpi_graph_get_next_endpoint()` calls `fwnode_get_next_child_node()`, which dispatches back to ACPI code, creating unnecessary indirection. - Solution: Call `acpi_get_next_subnode()` directly instead. No "Cc: stable(a)vger.kernel.org" tag, no "Fixes:" tag, no explicit bug report link. The message says "somewhat problematic," indicating an architectural issue rather than a critical bug. ### 2. CODE CHANGE ANALYSIS The diff shows 4 replacements in `acpi_graph_get_next_endpoint()`: - Line 1475: `fwnode_get_next_child_node(fwnode, port)` → `acpi_get_next_subnode(fwnode, port)` - Line 1493: `fwnode_get_next_child_node(port, prev)` → `acpi_get_next_subnode(port, prev)` - Line 1495: `fwnode_get_next_child_node(fwnode, port)` → `acpi_get_next_subnode(fwnode, port)` - Line 1499: `fwnode_get_next_child_node(port, NULL)` → `acpi_get_next_subnode(port, NULL)` Call chain: 1. `fwnode_get_next_child_node()` dispatches via `fwnode_call_ptr_op()` to the fwnode-specific implementation. 2. For ACPI fwnodes, it calls `acpi_get_next_present_subnode()` (registered at line 1747). 3. `acpi_get_next_present_subnode()` filters non-present device nodes and calls `acpi_get_next_subnode()`. Why the change is safe: - Graph endpoints are ACPI data nodes (checked by `is_acpi_graph_node()` at line 1448: `is_acpi_data_node(fwnode)`). - `acpi_get_next_present_subnode()` only filters non-present device nodes (lines 1407-1408), not data nodes. - Therefore, for graph endpoints, `acpi_get_next_subnode()` and `fwnode_get_next_child_node()` behave the same. ### 3. CLASSIFICATION This is a bug fix addressing an architectural issue: - Removes unnecessary indirection in ACPI-specific code. - Avoids a circular dependency pattern (ACPI → generic → ACPI). - Functionally equivalent for graph endpoints. Not a feature addition, not a new API, not a refactor. ### 4. SCOPE AND RISK ASSESSMENT - Scope: 4 lines changed in one function in one file. - Risk: Very low — same behavior for graph endpoints, cleaner architecture. - Complexity: Low — direct function call replacement. ### 5. USER IMPACT - Who is affected: Users of ACPI graph endpoints (e.g., camera/media drivers, device tree-like ACPI usage). - Severity: Low — architectural improvement, not a visible bug fix. - Likelihood: The "somewhat problematic" wording suggests no immediate user-visible issue. ### 6. STABILITY INDICATORS - Reviewed-by: Laurent Pinchart, Jonathan Cameron - Signed-off-by: Rafael J. Wysocki (ACPI maintainer) - No "Tested-by:" tags - Commit date: October 1, 2025 (recent) ### 7. DEPENDENCY CHECK - `acpi_get_next_subnode()` exists in the same file and has been present for years. - No external dependencies introduced. - Should apply cleanly to stable trees that have this code. ### 8. HISTORICAL CONTEXT Related commits: - `79389a83bc388`: Introduced `acpi_graph_get_next_endpoint()` with `fwnode_get_next_child_node()` calls. - `48698e6cf44c3`: Introduced `acpi_get_next_present_subnode()` to filter non-present devices. - `5d010473cdeaa` (this commit): Removes the indirection. The pattern existed since the function was introduced; this commit cleans it up. ### 9. STABLE KERNEL CRITERIA EVALUATION - Obviously correct: Yes — direct call instead of indirection. - Fixes a real bug: Yes — architectural issue that could cause problems. - Important issue: Moderate — architectural improvement, not a critical bug. - Small and contained: Yes — 4 lines, single function. - No new features: Yes — same behavior, cleaner code. - Applies cleanly: Yes — should apply without conflicts. ### 10. RISK VS BENEFIT Benefits: - Removes unnecessary indirection. - Avoids circular dependency pattern. - Improves code clarity. - No functional change for graph endpoints. Risks: - Very low — functionally equivalent change. - No new code paths or logic changes. ### 11. CONCERNS AND CONSIDERATIONS - No "Cc: stable" tag, but that alone doesn't disqualify. - Recent commit (Oct 2025) — hasn't been in mainline long. - No explicit bug report or user complaint mentioned. - Architectural improvement rather than a critical fix. ### CONCLUSION This is a small, correct fix that removes unnecessary indirection in ACPI code. It fixes an architectural issue and is functionally equivalent for graph endpoints. It meets stable kernel criteria: correct, fixes a real issue, small scope, no new features, and should apply cleanly. However, it's an architectural improvement rather than a critical bug fix, and there's no explicit stable tag or user-visible bug report. The "somewhat problematic" wording suggests it may not cause immediate visible problems. Given the conservative nature of stable trees and the lack of evidence of user-visible impact, this is borderline but leans toward acceptable for stable backporting due to its correctness, small scope, and architectural benefit. **YES** drivers/acpi/property.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/acpi/property.c b/drivers/acpi/property.c index 43d5e457814e1..76158b1399029 100644 --- a/drivers/acpi/property.c +++ b/drivers/acpi/property.c @@ -1472,7 +1472,7 @@ static struct fwnode_handle *acpi_graph_get_next_endpoint( if (!prev) { do { - port = fwnode_get_next_child_node(fwnode, port); + port = acpi_get_next_subnode(fwnode, port); /* * The names of the port nodes begin with "port@" * followed by the number of the port node and they also @@ -1490,13 +1490,13 @@ static struct fwnode_handle *acpi_graph_get_next_endpoint( if (!port) return NULL; - endpoint = fwnode_get_next_child_node(port, prev); + endpoint = acpi_get_next_subnode(port, prev); while (!endpoint) { - port = fwnode_get_next_child_node(fwnode, port); + port = acpi_get_next_subnode(fwnode, port); if (!port) break; if (is_acpi_graph_node(port, "port")) - endpoint = fwnode_get_next_child_node(port, NULL); + endpoint = acpi_get_next_subnode(port, NULL); } /* -- 2.51.0

1 week, 1 day

1
6
0 0

[PATCH net] net: phy: marvell-88q2xxx: Fix clamped value in mv88q2xxx_hwmon_write

by Thorsten Blum

The local variable 'val' was never clamped to -75000 or 180000 because the return value of clamp_val() was not used. Fix this by assigning the clamped value back to 'val', and use clamp() instead of clamp_val(). Cc: stable(a)vger.kernel.org Fixes: a557a92e6881 ("net: phy: marvell-88q2xxx: add support for temperature sensor") Signed-off-by: Thorsten Blum <thorsten.blum(a)linux.dev> --- drivers/net/phy/marvell-88q2xxx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/net/phy/marvell-88q2xxx.c b/drivers/net/phy/marvell-88q2xxx.c index f3d83b04c953..201dee1a1698 100644 --- a/drivers/net/phy/marvell-88q2xxx.c +++ b/drivers/net/phy/marvell-88q2xxx.c @@ -698,7 +698,7 @@ static int mv88q2xxx_hwmon_write(struct device *dev, switch (attr) { case hwmon_temp_max: - clamp_val(val, -75000, 180000); + val = clamp(val, -75000, 180000); val = (val / 1000) + 75; val = FIELD_PREP(MDIO_MMD_PCS_MV_TEMP_SENSOR3_INT_THRESH_MASK, val); -- Thorsten Blum <thorsten.blum(a)linux.dev> GPG: 1D60 735E 8AEF 3BE4 73B6 9D84 7336 78FD 8DFE EAD4

1 week, 1 day

4
3
0 0

[PATCH 1/1] phy: tegra: xusb: Fix UTMI AO sleepwalk trigger programming sequence

by Wayne Chang

From: Haotien Hsu <haotienh(a)nvidia.com> The UTMIP sleepwalk programming sequence requires asserting both LINEVAL_WALK_EN and WAKE_WALK_EN when enabling the sleepwalk logic. However, the current code mistakenly cleared WAKE_WALK_EN, which prevents the sleepwalk trigger from operating correctly. Fix this by asserting WAKE_WALK_EN together with LINEVAL_WALK_EN. Fixes: 1f9cab6cc20c ("phy: tegra: xusb: Add wake/sleepwalk for Tegra186") Cc: stable(a)vger.kernel.org Signed-off-by: Haotien Hsu <haotienh(a)nvidia.com> Signed-off-by: Wayne Chang <waynec(a)nvidia.com> --- drivers/phy/tegra/xusb-tegra186.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/phy/tegra/xusb-tegra186.c b/drivers/phy/tegra/xusb-tegra186.c index e818f6c3980e..b2a76710c0c4 100644 --- a/drivers/phy/tegra/xusb-tegra186.c +++ b/drivers/phy/tegra/xusb-tegra186.c @@ -401,8 +401,7 @@ static int tegra186_utmi_enable_phy_sleepwalk(struct tegra_xusb_lane *lane, /* enable the trigger of the sleepwalk logic */ value = ao_readl(priv, XUSB_AO_UTMIP_SLEEPWALK_CFG(index)); - value |= LINEVAL_WALK_EN; - value &= ~WAKE_WALK_EN; + value |= LINEVAL_WALK_EN | WAKE_WALK_EN; ao_writel(priv, value, XUSB_AO_UTMIP_SLEEPWALK_CFG(index)); /* reset the walk pointer and clear the alarm of the sleepwalk logic, -- 2.25.1

1 week, 1 day

2
2
0 0

[PATCH RESEND] drm: mipi-dsi: Fix an API misuse in mipi_dsi_device_register_full()

by Haoxiang Li

mipi_dsi_device_alloc() calls device_initialize() to initialize value "&dsi->dev". Thus "dsi" should be freed using put_device() in error handling path. Fixes: 068a00233969 ("drm: Add MIPI DSI bus support") Cc: stable(a)vger.kernel.org Signed-off-by: Haoxiang Li <haoxiang_li2024(a)163.com> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov(a)oss.qualcomm.com> --- drivers/gpu/drm/drm_mipi_dsi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/drm_mipi_dsi.c b/drivers/gpu/drm/drm_mipi_dsi.c index a712e177b350..e47950009126 100644 --- a/drivers/gpu/drm/drm_mipi_dsi.c +++ b/drivers/gpu/drm/drm_mipi_dsi.c @@ -233,7 +233,7 @@ mipi_dsi_device_register_full(struct mipi_dsi_host *host, ret = mipi_dsi_device_add(dsi); if (ret) { dev_err(host->dev, "failed to add DSI device %d\n", ret); - kfree(dsi); + put_device(&dsi->dev); return ERR_PTR(ret); } -- 2.25.1

1 week, 1 day

1
0
0 0

[PATCH 6.6] net: dsa: sja1105: fix kasan out-of-bounds warning in sja1105_table_delete_entry()

by Chen Yu

From: Vladimir Oltean <vladimir.oltean(a)nxp.com> [ Upstream commit 5f2b28b79d2d1946ee36ad8b3dc0066f73c90481 ] There are actually 2 problems: - deleting the last element doesn't require the memmove of elements [i + 1, end) over it. Actually, element i+1 is out of bounds. - The memmove itself should move size - i - 1 elements, because the last element is out of bounds. The out-of-bounds element still remains out of bounds after being accessed, so the problem is only that we touch it, not that it becomes in active use. But I suppose it can lead to issues if the out-of-bounds element is part of an unmapped page. Fixes: 6666cebc5e30 ("net: dsa: sja1105: Add support for VLAN operations") Signed-off-by: Vladimir Oltean <vladimir.oltean(a)nxp.com> Reviewed-by: Simon Horman <horms(a)kernel.org> Link: https://patch.msgid.link/20250318115716.2124395-4-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Signed-off-by: Chen Yu <xnguchen(a)sina.cn> --- drivers/net/dsa/sja1105/sja1105_static_config.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/net/dsa/sja1105/sja1105_static_config.c b/drivers/net/dsa/sja1105/sja1105_static_config.c index baba204ad62f..2ac91fe2a79b 100644 --- a/drivers/net/dsa/sja1105/sja1105_static_config.c +++ b/drivers/net/dsa/sja1105/sja1105_static_config.c @@ -1921,8 +1921,10 @@ int sja1105_table_delete_entry(struct sja1105_table *table, int i) if (i > table->entry_count) return -ERANGE; - memmove(entries + i * entry_size, entries + (i + 1) * entry_size, - (table->entry_count - i) * entry_size); + if (i + 1 < table->entry_count) { + memmove(entries + i * entry_size, entries + (i + 1) * entry_size, + (table->entry_count - i - 1) * entry_size); + } table->entry_count--; -- 2.17.1

1 week, 1 day

1
0
0 0

[GIT PULL] virtio,vhost: fixes, cleanups

by Michael S. Tsirkin

get_user/put_user change didn't spend time in next and seems a bit too risky to rush. I'm keeping it in my tree and we'll get it in the next cycle. The following changes since commit ac3fd01e4c1efce8f2c054cdeb2ddd2fc0fb150d: Linux 6.18-rc7 (2025-11-23 14:53:16 -0800) are available in the Git repository at: https://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost.git tags/for_linus for you to fetch changes up to 205dd7a5d6ad6f4c8e8fcd3c3b95a7c0e7067fee: virtio_pci: drop kernel.h (2025-11-30 18:02:43 -0500) ---------------------------------------------------------------- virtio,vhost: fixes, cleanups Just a bunch of fixes and cleanups, mostly very simple. Several features are merged through net-next this time around. Signed-off-by: Michael S. Tsirkin <mst(a)redhat.com> ---------------------------------------------------------------- Alok Tiwari (3): virtio_vdpa: fix misleading return in void function vdpa/mlx5: Fix incorrect error code reporting in query_virtqueues vdpa/pds: use %pe for ERR_PTR() in event handler registration Kriish Sharma (1): virtio: fix kernel-doc for mapping/free_coherent functions Marco Crivellari (2): virtio_balloon: add WQ_PERCPU to alloc_workqueue users vduse: add WQ_PERCPU to alloc_workqueue users Miaoqian Lin (1): virtio: vdpa: Fix reference count leak in octep_sriov_enable() Michael S. Tsirkin (11): virtio: fix typo in virtio_device_ready() comment virtio: fix whitespace in virtio_config_ops virtio: fix grammar in virtio_queue_info docs virtio: fix grammar in virtio_map_ops docs virtio: standardize Returns documentation style virtio: fix virtqueue_set_affinity() docs virtio: fix map ops comment virtio: clean up features qword/dword terms vhost/test: add test specific macro for features vhost: switch to arrays of feature bits virtio_pci: drop kernel.h Mike Christie (1): vhost: Fix kthread worker cgroup failure handling drivers/vdpa/mlx5/net/mlx5_vnet.c | 2 +- drivers/vdpa/octeon_ep/octep_vdpa_main.c | 1 + drivers/vdpa/pds/vdpa_dev.c | 2 +- drivers/vdpa/vdpa_user/vduse_dev.c | 3 ++- drivers/vhost/net.c | 29 +++++++++++----------- drivers/vhost/scsi.c | 9 ++++--- drivers/vhost/test.c | 10 ++++++-- drivers/vhost/vhost.c | 4 ++- drivers/vhost/vhost.h | 42 ++++++++++++++++++++++++++------ drivers/vhost/vsock.c | 10 +++++--- drivers/virtio/virtio.c | 12 ++++----- drivers/virtio/virtio_balloon.c | 3 ++- drivers/virtio/virtio_debug.c | 10 ++++---- drivers/virtio/virtio_pci_modern_dev.c | 6 ++--- drivers/virtio/virtio_ring.c | 7 +++--- drivers/virtio/virtio_vdpa.c | 2 +- include/linux/virtio.h | 2 +- include/linux/virtio_config.h | 24 +++++++++--------- include/linux/virtio_features.h | 29 +++++++++++----------- include/linux/virtio_pci_modern.h | 8 +++--- include/uapi/linux/virtio_pci.h | 2 +- 21 files changed, 131 insertions(+), 86 deletions(-)

1 week, 1 day

2
1
0 0

[PATCH v2 1/9] crypto: virtio: Add spinlock protection with virtqueue notification

by Bibo Mao

When VM boots with one virtio-crypto PCI device and builtin backend, run openssl benchmark command with multiple processes, such as openssl speed -evp aes-128-cbc -engine afalg -seconds 10 -multi 32 openssl processes will hangup and there is error reported like this: virtio_crypto virtio0: dataq.0:id 3 is not a head! It seems that the data virtqueue need protection when it is handled for virtio done notification. If the spinlock protection is added in virtcrypto_done_task(), openssl benchmark with multiple processes works well. Fixes: fed93fb62e05 ("crypto: virtio - Handle dataq logic with tasklet") Cc: stable(a)vger.kernel.org Signed-off-by: Bibo Mao <maobibo(a)loongson.cn> --- drivers/crypto/virtio/virtio_crypto_core.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/crypto/virtio/virtio_crypto_core.c b/drivers/crypto/virtio/virtio_crypto_core.c index 3d241446099c..ccc6b5c1b24b 100644 --- a/drivers/crypto/virtio/virtio_crypto_core.c +++ b/drivers/crypto/virtio/virtio_crypto_core.c @@ -75,15 +75,20 @@ static void virtcrypto_done_task(unsigned long data) struct data_queue *data_vq = (struct data_queue *)data; struct virtqueue *vq = data_vq->vq; struct virtio_crypto_request *vc_req; + unsigned long flags; unsigned int len; + spin_lock_irqsave(&data_vq->lock, flags); do { virtqueue_disable_cb(vq); while ((vc_req = virtqueue_get_buf(vq, &len)) != NULL) { + spin_unlock_irqrestore(&data_vq->lock, flags); if (vc_req->alg_cb) vc_req->alg_cb(vc_req, len); + spin_lock_irqsave(&data_vq->lock, flags); } } while (!virtqueue_enable_cb(vq)); + spin_unlock_irqrestore(&data_vq->lock, flags); } static void virtcrypto_dataq_callback(struct virtqueue *vq) -- 2.39.3

1 week, 1 day

2
2
0 0

[PATCH V2] mm/slab: introduce kvfree_rcu_barrier_on_cache() for cache destruction

by Harry Yoo

Currently, kvfree_rcu_barrier() flushes RCU sheaves across all slab caches when a cache is destroyed. This is unnecessary; only the RCU sheaves belonging to the cache being destroyed need to be flushed. As suggested by Vlastimil Babka, introduce a weaker form of kvfree_rcu_barrier() that operates on a specific slab cache. Factor out flush_rcu_sheaves_on_cache() from flush_all_rcu_sheaves() and call it from flush_all_rcu_sheaves() and kvfree_rcu_barrier_on_cache(). Call kvfree_rcu_barrier_on_cache() instead of kvfree_rcu_barrier() on cache destruction. The performance benefit is evaluated on a 12 core 24 threads AMD Ryzen 5900X machine (1 socket), by loading slub_kunit module. Before: Total calls: 19 Average latency (us): 18127 Total time (us): 344414 After: Total calls: 19 Average latency (us): 10066 Total time (us): 191264 Two performance regression have been reported: - stress module loader test's runtime increases by 50-60% (Daniel) - internal graphics test's runtime on Tegra23 increases by 35% (Jon) They are fixed by this change. Suggested-by: Vlastimil Babka <vbabka(a)suse.cz> Fixes: ec66e0d59952 ("slab: add sheaf support for batching kfree_rcu() operations") Cc: <stable(a)vger.kernel.org> Link: https://lore.kernel.org/linux-mm/1bda09da-93be-4737-aef0-d47f8c5c9301@suse.… Reported-and-tested-by: Daniel Gomez <da.gomez(a)samsung.com> Closes: https://lore.kernel.org/linux-mm/0406562e-2066-4cf8-9902-b2b0616dd742@kerne… Reported-and-tested-by: Jon Hunter <jonathanh(a)nvidia.com> Closes: https://lore.kernel.org/linux-mm/e988eff6-1287-425e-a06c-805af5bbf262@nvidi… Signed-off-by: Harry Yoo <harry.yoo(a)oracle.com> --- No code change, added proper tags and updated changelog. include/linux/slab.h | 5 ++++ mm/slab.h | 1 + mm/slab_common.c | 52 +++++++++++++++++++++++++++++------------ mm/slub.c | 55 ++++++++++++++++++++++++-------------------- 4 files changed, 73 insertions(+), 40 deletions(-) diff --git a/include/linux/slab.h b/include/linux/slab.h index cf443f064a66..937c93d44e8c 100644 --- a/include/linux/slab.h +++ b/include/linux/slab.h @@ -1149,6 +1149,10 @@ static inline void kvfree_rcu_barrier(void) { rcu_barrier(); } +static inline void kvfree_rcu_barrier_on_cache(struct kmem_cache *s) +{ + rcu_barrier(); +} static inline void kfree_rcu_scheduler_running(void) { } #else @@ -1156,6 +1160,7 @@ void kvfree_rcu_barrier(void); void kfree_rcu_scheduler_running(void); #endif +void kvfree_rcu_barrier_on_cache(struct kmem_cache *s); /** * kmalloc_size_roundup - Report allocation bucket size for the given size diff --git a/mm/slab.h b/mm/slab.h index f730e012553c..e767aa7e91b0 100644 --- a/mm/slab.h +++ b/mm/slab.h @@ -422,6 +422,7 @@ static inline bool is_kmalloc_normal(struct kmem_cache *s) bool __kfree_rcu_sheaf(struct kmem_cache *s, void *obj); void flush_all_rcu_sheaves(void); +void flush_rcu_sheaves_on_cache(struct kmem_cache *s); #define SLAB_CORE_FLAGS (SLAB_HWCACHE_ALIGN | SLAB_CACHE_DMA | \ SLAB_CACHE_DMA32 | SLAB_PANIC | \ diff --git a/mm/slab_common.c b/mm/slab_common.c index 84dfff4f7b1f..dd8a49d6f9cc 100644 --- a/mm/slab_common.c +++ b/mm/slab_common.c @@ -492,7 +492,7 @@ void kmem_cache_destroy(struct kmem_cache *s) return; /* in-flight kfree_rcu()'s may include objects from our cache */ - kvfree_rcu_barrier(); + kvfree_rcu_barrier_on_cache(s); if (IS_ENABLED(CONFIG_SLUB_RCU_DEBUG) && (s->flags & SLAB_TYPESAFE_BY_RCU)) { @@ -2038,25 +2038,13 @@ void kvfree_call_rcu(struct rcu_head *head, void *ptr) } EXPORT_SYMBOL_GPL(kvfree_call_rcu); -/** - * kvfree_rcu_barrier - Wait until all in-flight kvfree_rcu() complete. - * - * Note that a single argument of kvfree_rcu() call has a slow path that - * triggers synchronize_rcu() following by freeing a pointer. It is done - * before the return from the function. Therefore for any single-argument - * call that will result in a kfree() to a cache that is to be destroyed - * during module exit, it is developer's responsibility to ensure that all - * such calls have returned before the call to kmem_cache_destroy(). - */ -void kvfree_rcu_barrier(void) +static inline void __kvfree_rcu_barrier(void) { struct kfree_rcu_cpu_work *krwp; struct kfree_rcu_cpu *krcp; bool queued; int i, cpu; - flush_all_rcu_sheaves(); - /* * Firstly we detach objects and queue them over an RCU-batch * for all CPUs. Finally queued works are flushed for each CPU. @@ -2118,8 +2106,43 @@ void kvfree_rcu_barrier(void) } } } + +/** + * kvfree_rcu_barrier - Wait until all in-flight kvfree_rcu() complete. + * + * Note that a single argument of kvfree_rcu() call has a slow path that + * triggers synchronize_rcu() following by freeing a pointer. It is done + * before the return from the function. Therefore for any single-argument + * call that will result in a kfree() to a cache that is to be destroyed + * during module exit, it is developer's responsibility to ensure that all + * such calls have returned before the call to kmem_cache_destroy(). + */ +void kvfree_rcu_barrier(void) +{ + flush_all_rcu_sheaves(); + __kvfree_rcu_barrier(); +} EXPORT_SYMBOL_GPL(kvfree_rcu_barrier); +/** + * kvfree_rcu_barrier_on_cache - Wait for in-flight kvfree_rcu() calls on a + * specific slab cache. + * @s: slab cache to wait for + * + * See the description of kvfree_rcu_barrier() for details. + */ +void kvfree_rcu_barrier_on_cache(struct kmem_cache *s) +{ + if (s->cpu_sheaves) + flush_rcu_sheaves_on_cache(s); + /* + * TODO: Introduce a version of __kvfree_rcu_barrier() that works + * on a specific slab cache. + */ + __kvfree_rcu_barrier(); +} +EXPORT_SYMBOL_GPL(kvfree_rcu_barrier_on_cache); + static unsigned long kfree_rcu_shrink_count(struct shrinker *shrink, struct shrink_control *sc) { @@ -2215,4 +2238,3 @@ void __init kvfree_rcu_init(void) } #endif /* CONFIG_KVFREE_RCU_BATCHED */ - diff --git a/mm/slub.c b/mm/slub.c index 785e25a14999..7cec2220712b 100644 --- a/mm/slub.c +++ b/mm/slub.c @@ -4118,42 +4118,47 @@ static void flush_rcu_sheaf(struct work_struct *w) /* needed for kvfree_rcu_barrier() */ -void flush_all_rcu_sheaves(void) +void flush_rcu_sheaves_on_cache(struct kmem_cache *s) { struct slub_flush_work *sfw; - struct kmem_cache *s; unsigned int cpu; - cpus_read_lock(); - mutex_lock(&slab_mutex); + mutex_lock(&flush_lock); - list_for_each_entry(s, &slab_caches, list) { - if (!s->cpu_sheaves) - continue; + for_each_online_cpu(cpu) { + sfw = &per_cpu(slub_flush, cpu); - mutex_lock(&flush_lock); + /* + * we don't check if rcu_free sheaf exists - racing + * __kfree_rcu_sheaf() might have just removed it. + * by executing flush_rcu_sheaf() on the cpu we make + * sure the __kfree_rcu_sheaf() finished its call_rcu() + */ - for_each_online_cpu(cpu) { - sfw = &per_cpu(slub_flush, cpu); + INIT_WORK(&sfw->work, flush_rcu_sheaf); + sfw->s = s; + queue_work_on(cpu, flushwq, &sfw->work); + } - /* - * we don't check if rcu_free sheaf exists - racing - * __kfree_rcu_sheaf() might have just removed it. - * by executing flush_rcu_sheaf() on the cpu we make - * sure the __kfree_rcu_sheaf() finished its call_rcu() - */ + for_each_online_cpu(cpu) { + sfw = &per_cpu(slub_flush, cpu); + flush_work(&sfw->work); + } - INIT_WORK(&sfw->work, flush_rcu_sheaf); - sfw->s = s; - queue_work_on(cpu, flushwq, &sfw->work); - } + mutex_unlock(&flush_lock); +} - for_each_online_cpu(cpu) { - sfw = &per_cpu(slub_flush, cpu); - flush_work(&sfw->work); - } +void flush_all_rcu_sheaves(void) +{ + struct kmem_cache *s; + + cpus_read_lock(); + mutex_lock(&slab_mutex); - mutex_unlock(&flush_lock); + list_for_each_entry(s, &slab_caches, list) { + if (!s->cpu_sheaves) + continue; + flush_rcu_sheaves_on_cache(s); } mutex_unlock(&slab_mutex); -- 2.43.0

1 week, 1 day

3
5
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror