Linux-stable-mirror

linux-stable-mirror@lists.linaro.org

305 participants
124182 discussions

[PATCH can v2] can: gs_usb: gs_usb_receive_bulk_callback(): fix URB memory leak

by Marc Kleine-Budde

In gs_can_open(), the URBs for USB-in transfers are allocated, added to the parent->rx_submitted anchor and submitted. In the complete callback gs_usb_receive_bulk_callback(), the URB is processed and resubmitted. In gs_can_close() the URBs are freed by calling usb_kill_anchored_urbs(parent->rx_submitted). However, this does not take into account that the USB framework unanchors the URB before the close function is called. This means that once an in-URB has been completed, it is no longer anchored and is ultimately not released in gs_can_close(). Fix the memory leak by anchoring the URB in the gs_usb_receive_bulk_callback() to the parent->rx_submitted anchor. Fixes: d08e973a77d1 ("can: gs_usb: Added support for the GS_USB CAN devices") Cc: stable(a)vger.kernel.org Signed-off-by: Marc Kleine-Budde <mkl(a)pengutronix.de> --- Changes in v2: - correct "Fixes" tag - add stable(a)v.k.o on Cc - Link to v1: https://patch.msgid.link/20251225-gs_usb-fix-memory-leak-v1-1-a7b09e70d01d@… --- drivers/net/can/usb/gs_usb.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/net/can/usb/gs_usb.c b/drivers/net/can/usb/gs_usb.c index a0233e550a5a..d093babbc320 100644 --- a/drivers/net/can/usb/gs_usb.c +++ b/drivers/net/can/usb/gs_usb.c @@ -751,6 +751,8 @@ static void gs_usb_receive_bulk_callback(struct urb *urb) hf, parent->hf_size_rx, gs_usb_receive_bulk_callback, parent); + usb_anchor_urb(urb, &parent->rx_submitted); + rc = usb_submit_urb(urb, GFP_ATOMIC); /* USB failure take down all interfaces */ --- base-commit: 1806d210e5a8f431ad4711766ae4a333d407d972 change-id: 20251225-gs_usb-fix-memory-leak-062c24898cc9 Best regards, -- Marc Kleine-Budde <mkl(a)pengutronix.de>

3 days, 2 hours

[PATCH] pinctrl: meson: mark the GPIO controller as sleeping

by Bartosz Golaszewski

The GPIO controller is configured as non-sleeping but it uses generic pinctrl helpers which use a mutex for synchronization. This can cause the following lockdep splat with shared GPIOs enabled on boards which have multiple devices using the same GPIO: BUG: sleeping function called from invalid context at kernel/locking/mutex.c:591 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 142, name: kworker/u25:3 preempt_count: 1, expected: 0 RCU nest depth: 0, expected: 0 INFO: lockdep is turned off. irq event stamp: 46379 hardirqs last enabled at (46379): [<ffff8000813acb24>] _raw_spin_unlock_irqrestore+0x74/0x78 hardirqs last disabled at (46378): [<ffff8000813abf38>] _raw_spin_lock_irqsave+0x84/0x88 softirqs last enabled at (46330): [<ffff8000800c71b4>] handle_softirqs+0x4c4/0x4dc softirqs last disabled at (46295): [<ffff800080010674>] __do_softirq+0x14/0x20 CPU: 1 UID: 0 PID: 142 Comm: kworker/u25:3 Tainted: G C 6.19.0-rc4-next-20260105+ #11963 PREEMPT Tainted: [C]=CRAP Hardware name: Khadas VIM3 (DT) Workqueue: events_unbound deferred_probe_work_func Call trace: show_stack+0x18/0x24 (C) dump_stack_lvl+0x90/0xd0 dump_stack+0x18/0x24 __might_resched+0x144/0x248 __might_sleep+0x48/0x98 __mutex_lock+0x5c/0x894 mutex_lock_nested+0x24/0x30 pinctrl_get_device_gpio_range+0x44/0x128 pinctrl_gpio_set_config+0x40/0xdc gpiochip_generic_config+0x28/0x3c gpio_do_set_config+0xa8/0x194 gpiod_set_config+0x34/0xfc gpio_shared_proxy_set_config+0x6c/0xfc [gpio_shared_proxy] gpio_do_set_config+0xa8/0x194 gpiod_set_transitory+0x4c/0xf0 gpiod_configure_flags+0xa4/0x480 gpiod_find_and_request+0x1a0/0x574 gpiod_get_index+0x58/0x84 devm_gpiod_get_index+0x20/0xb4 devm_gpiod_get+0x18/0x24 mmc_pwrseq_emmc_probe+0x40/0xb8 platform_probe+0x5c/0xac really_probe+0xbc/0x298 __driver_probe_device+0x78/0x12c driver_probe_device+0xdc/0x164 __device_attach_driver+0xb8/0x138 bus_for_each_drv+0x80/0xdc __device_attach+0xa8/0x1b0 Fixes: 6ac730951104 ("pinctrl: add driver for Amlogic Meson SoCs") Cc: stable(a)vger.kernel.org Reported-by: Marek Szyprowski <m.szyprowski(a)samsung.com> Closes: https://lore.kernel.org/all/00107523-7737-4b92-a785-14ce4e93b8cb@samsung.co… Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski(a)oss.qualcomm.com> --- drivers/pinctrl/meson/pinctrl-meson.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/pinctrl/meson/pinctrl-meson.c b/drivers/pinctrl/meson/pinctrl-meson.c index 18295b15ecd9..4507dc8b5563 100644 --- a/drivers/pinctrl/meson/pinctrl-meson.c +++ b/drivers/pinctrl/meson/pinctrl-meson.c @@ -619,7 +619,7 @@ static int meson_gpiolib_register(struct meson_pinctrl *pc) pc->chip.set = meson_gpio_set; pc->chip.base = -1; pc->chip.ngpio = pc->data->num_pins; - pc->chip.can_sleep = false; + pc->chip.can_sleep = true; ret = gpiochip_add_data(&pc->chip, pc); if (ret) { -- 2.47.3

3 days, 2 hours

[PATCH] drm/xe: fix WQ_MEM_RECLAIM passed as max_active to alloc_workqueue()

by Marco Crivellari

Workqueue xe-ggtt-wq has been allocated using WQ_MEM_RECLAIM, but the flag has been passed as 3rd parameter (max_active) instead of 2nd (flags) creating the workqueue as per-cpu with max_active = 8 (the WQ_MEM_RECLAIM value). So change this by set WQ_MEM_RECLAIM as the 2nd parameter with a default max_active. Fixes: 60df57e496e4 ("drm/xe: Mark GGTT work queue with WQ_MEM_RECLAIM") Cc: stable(a)vger.kernel.org Signed-off-by: Marco Crivellari <marco.crivellari(a)suse.com> --- drivers/gpu/drm/xe/xe_ggtt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_ggtt.c b/drivers/gpu/drm/xe/xe_ggtt.c index ef481b334af4..793d7324a395 100644 --- a/drivers/gpu/drm/xe/xe_ggtt.c +++ b/drivers/gpu/drm/xe/xe_ggtt.c @@ -322,7 +322,7 @@ int xe_ggtt_init_early(struct xe_ggtt *ggtt) else ggtt->pt_ops = &xelp_pt_ops; - ggtt->wq = alloc_workqueue("xe-ggtt-wq", 0, WQ_MEM_RECLAIM); + ggtt->wq = alloc_workqueue("xe-ggtt-wq", WQ_MEM_RECLAIM, 0); if (!ggtt->wq) return -ENOMEM; -- 2.52.0

3 days, 3 hours

[PATCH v2 00/12] drm/bridge: convert users of of_drm_find_bridge(), part 2

by Luca Ceresoli

This series converts all DRM bridge drivers (*) from the now deprecated of_drm_find_bridge() to its replacement of_drm_find_and_get_bridge() which allows correct bridge refcounting. It also converts per-driver "next_bridge" pointers to the unified drm_bridge::next_bridge which puts the reference automatically on bridge deallocation. This is part of the work to support hotplug of DRM bridges. The grand plan was discussed in [0]. Here's the work breakdown (➜ marks the current series): 1. ➜ add refcounting to DRM bridges struct drm_bridge, based on devm_drm_bridge_alloc() A. ✔ add new alloc API and refcounting (v6.16) B. ✔ convert all bridge drivers to new API (v6.17) C. ✔ kunit tests (v6.17) D. ✔ add get/put to drm_bridge_add/remove() + attach/detach() and warn on old allocation pattern (v6.17) E. ➜ add get/put on drm_bridge accessors 1. ✔ drm_bridge_chain_get_first_bridge(), add cleanup action (v6.18) 2. ✔ drm_bridge_get_prev_bridge() (v6.18) 3. ✔ drm_bridge_get_next_bridge() (v6.19) 4. ✔ drm_for_each_bridge_in_chain() (v6.19) 5. ✔ drm_bridge_connector_init (v6.19) 6. … protect encoder bridge chain with a mutex 7. ➜ of_drm_find_bridge a. ✔… add of_drm_get_bridge(), convert basic direct users (v6.20?, one driver still pending) b. ➜ convert direct of_drm_get_bridge() users, part 2 c. … convert direct of_drm_get_bridge() users, part 3 d. convert direct of_drm_get_bridge() users, part 4 e. convert bridge-only drm_of_find_panel_or_bridge() users 8. drm_of_find_panel_or_bridge, *_of_get_bridge 9. ✔ enforce drm_bridge_add before drm_bridge_attach (v6.19) F. ✔ debugfs improvements 1. ✔ add top-level 'bridges' file (v6.16) 2. ✔ show refcount and list lingering bridges (v6.19) 2. … handle gracefully atomic updates during bridge removal A. ✔ Add drm_dev_enter/exit() to protect device resources (v6.20?) B. … protect private_obj removal from list 3. … DSI host-device driver interaction 4. ✔ removing the need for the "always-disconnected" connector 5. finish the hotplug bridge work, moving code to the core and potentially removing the hotplug-bridge itself (this needs to be clarified as points 1-3 are developed) [0] https://lore.kernel.org/lkml/20250206-hotplug-drm-bridge-v6-0-9d6f2c9c3058@… This work is a continuation of the work to correctly handle bridge refcounting for existing of_drm_find_bridge(). The ground work is in: - commit 293a8fd7721a ("drm/bridge: add of_drm_find_and_get_bridge()") - commit 9da0e06abda8 ("drm/bridge: deprecate of_drm_find_bridge()") - commit 3fdeae134ba9 ("drm/bridge: add next_bridge pointer to struct drm_bridge") The whole conversion is split in multiple series to make the review process a bit smoother. Parts 3 and 4 are converting non-bridge drivers (mostly encoders). (*) One bridge driver (synopsys/dw-hdmi) is converted in another series, together with its (non-bridge) users. Additionally this series converts drm_of_panel_bridge_remove() which is a special case, and has a bugfix for it too. Signed-off-by: Luca Ceresoli <luca.ceresoli(a)bootlin.com> --- Changes in v2: - Fix bug in samsung-dsim code, patch 10; update patch 12 on top of those changes - Fix in imx8qxp-ldb code, patch 9 - Link to v1: https://lore.kernel.org/r/20260107-drm-bridge-alloc-getput-drm_of_find_brid… --- Luca Ceresoli (12): drm: of: drm_of_panel_bridge_remove(): fix device_node leak drm: of: drm_of_panel_bridge_remove(): convert to of_drm_find_and_get_bridge() drm/bridge: sii902x: convert to of_drm_find_and_get_bridge() drm/bridge: thc63lvd1024: convert to of_drm_find_and_get_bridge() drm/bridge: tfp410: convert to of_drm_find_and_get_bridge() drm/bridge: tpd12s015: convert to of_drm_find_and_get_bridge() drm/bridge: lt8912b: convert to of_drm_find_and_get_bridge() drm/bridge: imx8mp-hdmi-pvi: convert to of_drm_find_and_get_bridge() drm/bridge: imx8qxp-ldb: convert to of_drm_find_and_get_bridge() drm/bridge: samsung-dsim: samsung_dsim_host_attach: use a temporary variable for the next bridge drm/bridge: samsung-dsim: samsung_dsim_host_attach: don't use the bridge pointer as an error indicator drm/bridge: samsung-dsim: samsung_dsim_host_attach: convert to of_drm_find_and_get_bridge() drivers/gpu/drm/bridge/imx/imx8mp-hdmi-pvi.c | 15 ++++++----- drivers/gpu/drm/bridge/imx/imx8qxp-ldb.c | 12 ++++++++- drivers/gpu/drm/bridge/lontium-lt8912b.c | 31 +++++++++++------------ drivers/gpu/drm/bridge/samsung-dsim.c | 37 +++++++++++++++++++--------- drivers/gpu/drm/bridge/sii902x.c | 7 +++--- drivers/gpu/drm/bridge/thc63lvd1024.c | 7 +++--- drivers/gpu/drm/bridge/ti-tfp410.c | 27 ++++++++++---------- drivers/gpu/drm/bridge/ti-tpd12s015.c | 8 +++--- include/drm/bridge/samsung-dsim.h | 1 - include/drm/drm_of.h | 6 ++++- 10 files changed, 86 insertions(+), 65 deletions(-) --- base-commit: 814b88f91b121246b35be9c519b537041fe3d178 change-id: 20251223-drm-bridge-alloc-getput-drm_of_find_bridge-2-12c6bbcb6896 Best regards, -- Luca Ceresoli <luca.ceresoli(a)bootlin.com>

3 days, 4 hours

[PATCH v3] usb: xhci: Fix memory leak in xhci_disable_slot()

by Zilin Guan

xhci_alloc_command() allocates a command structure and, when the second argument is true, also allocates a completion structure. Currently, the error handling path in xhci_disable_slot() only frees the command structure using kfree(), causing the completion structure to leak. Use xhci_free_command() instead of kfree(). xhci_free_command() correctly frees both the command structure and the associated completion structure. Since the command structure is allocated with zero-initialization, command->in_ctx is NULL and will not be erroneously freed by xhci_free_command(). This bug was found using an experimental static analysis tool we are developing. The tool is based on the LLVM framework and is specifically designed to detect memory management issues. It is currently under active development and not yet publicly available, but we plan to open-source it after our research is published. The bug was originally detected on v6.13-rc1 using our static analysis tool, and we have verified that the issue persists in the latest mainline kernel. We performed build testing on x86_64 with allyesconfig using GCC=11.4.0. Since triggering these error paths in xhci_disable_slot() requires specific hardware conditions or abnormal state, we were unable to construct a test case to reliably trigger these specific error paths at runtime. Fixes: 7faac1953ed1 ("xhci: avoid race between disable slot command and host runtime suspend") CC: stable(a)vger.kernel.org Signed-off-by: Zilin Guan <zilin(a)seu.edu.cn> --- Changes in v3: - Update the commit message to reflect that the analysis applies to the mainline kernel. Changes in v2: - Add detailed information required by researcher guidelines. - Clarify the safety of using xhci_free_command() in this context. - Correct the Fixes tag to point to the commit that introduced this issue. drivers/usb/host/xhci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c index 02c9bfe21ae2..f0beed054954 100644 --- a/drivers/usb/host/xhci.c +++ b/drivers/usb/host/xhci.c @@ -4137,7 +4137,7 @@ int xhci_disable_slot(struct xhci_hcd *xhci, u32 slot_id) if (state == 0xffffffff || (xhci->xhc_state & XHCI_STATE_DYING) || (xhci->xhc_state & XHCI_STATE_HALTED)) { spin_unlock_irqrestore(&xhci->lock, flags); - kfree(command); + xhci_free_command(xhci, command); return -ENODEV; } @@ -4145,7 +4145,7 @@ int xhci_disable_slot(struct xhci_hcd *xhci, u32 slot_id) slot_id); if (ret) { spin_unlock_irqrestore(&xhci->lock, flags); - kfree(command); + xhci_free_command(xhci, command); return ret; } xhci_ring_cmd_db(xhci); -- 2.34.1

3 days, 6 hours

[PATCH 1/2] migrate: Correct lock ordering for hugetlb file folios

by Matthew Wilcox (Oracle)

Syzbot has found a deadlock (analyzed by Lance Yang): 1) Task (5749): Holds folio_lock, then tries to acquire i_mmap_rwsem(read lock). 2) Task (5754): Holds i_mmap_rwsem(write lock), then tries to acquire folio_lock. migrate_pages() -> migrate_hugetlbs() -> unmap_and_move_huge_page() <- Takes folio_lock! -> remove_migration_ptes() -> __rmap_walk_file() -> i_mmap_lock_read() <- Waits for i_mmap_rwsem(read lock)! hugetlbfs_fallocate() -> hugetlbfs_punch_hole() <- Takes i_mmap_rwsem(write lock)! -> hugetlbfs_zero_partial_page() -> filemap_lock_hugetlb_folio() -> filemap_lock_folio() -> __filemap_get_folio <- Waits for folio_lock! The migration path is the one taking locks in the wrong order according to the documentation at the top of mm/rmap.c. So expand the scope of the existing i_mmap_lock to cover the calls to remove_migration_ptes() too. This is (mostly) how it used to be after commit c0d0381ade79. That was removed by 336bf30eb765 for both file & anon hugetlb pages when it should only have been removed for anon hugetlb pages. Fixes: 336bf30eb765 (hugetlbfs: fix anon huge page migration race) Reported-by: syzbot+2d9c96466c978346b55f(a)syzkaller.appspotmail.com Link: https://lore.kernel.org/all/68e9715a.050a0220.1186a4.000d.GAE@google.com Debugged-by: Lance Yang <lance.yang(a)linux.dev> Signed-off-by: Matthew Wilcox (Oracle) <willy(a)infradead.org> Cc: stable(a)vger.kernel.org --- mm/migrate.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 5169f9717f60..4688b9e38cd2 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1458,6 +1458,7 @@ static int unmap_and_move_huge_page(new_folio_t get_new_folio, int page_was_mapped = 0; struct anon_vma *anon_vma = NULL; struct address_space *mapping = NULL; + enum ttu_flags ttu = 0; if (folio_ref_count(src) == 1) { /* page was freed from under us. So we are done. */ @@ -1498,8 +1499,6 @@ static int unmap_and_move_huge_page(new_folio_t get_new_folio, goto put_anon; if (folio_mapped(src)) { - enum ttu_flags ttu = 0; - if (!folio_test_anon(src)) { /* * In shared mappings, try_to_unmap could potentially @@ -1516,16 +1515,17 @@ static int unmap_and_move_huge_page(new_folio_t get_new_folio, try_to_migrate(src, ttu); page_was_mapped = 1; - - if (ttu & TTU_RMAP_LOCKED) - i_mmap_unlock_write(mapping); } if (!folio_mapped(src)) rc = move_to_new_folio(dst, src, mode); if (page_was_mapped) - remove_migration_ptes(src, !rc ? dst : src, 0); + remove_migration_ptes(src, !rc ? dst : src, + ttu ? RMP_LOCKED : 0); + + if (ttu & TTU_RMAP_LOCKED) + i_mmap_unlock_write(mapping); unlock_put_anon: folio_unlock(dst); -- 2.47.3

3 days, 7 hours

[PATCH v2] usb: xhci: Fix memory leak in xhci_disable_slot()

by Zilin Guan

xhci_alloc_command() allocates a command structure and, when the second argument is true, also allocates a completion structure. Currently, the error handling path in xhci_disable_slot() only frees the command structure using kfree(), causing the completion structure to leak. Use xhci_free_command() instead of kfree(). xhci_free_command() correctly frees both the command structure and the associated completion structure. Since the command structure is allocated with zero-initialization, command->in_ctx is NULL and will not be erroneously freed by xhci_free_command(). This bug was found using an experimental static analysis tool we are developing. The tool is based on the LLVM framework and is specifically designed to detect memory management issues. It is currently under active development and not yet publicly available, but we plan to open-source it after our research is published. The analysis was performed on Linux kernel v6.13-rc1. We performed build testing on x86_64 with allyesconfig using GCC=11.4.0. Since triggering these error paths in xhci_disable_slot() requires specific hardware conditions or abnormal state, we were unable to construct a test case to reliably trigger these specific error paths at runtime. Fixes: 7faac1953ed1 ("xhci: avoid race between disable slot command and host runtime suspend") CC: stable(a)vger.kernel.org Signed-off-by: Zilin Guan <zilin(a)seu.edu.cn> --- Changes in v2: - Add detailed information required by researcher guidelines. - Clarify the safety of using xhci_free_command() in this context. - Correct the Fixes tag to point to the commit that introduced this issue. drivers/usb/host/xhci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c index 02c9bfe21ae2..f0beed054954 100644 --- a/drivers/usb/host/xhci.c +++ b/drivers/usb/host/xhci.c @@ -4137,7 +4137,7 @@ int xhci_disable_slot(struct xhci_hcd *xhci, u32 slot_id) if (state == 0xffffffff || (xhci->xhc_state & XHCI_STATE_DYING) || (xhci->xhc_state & XHCI_STATE_HALTED)) { spin_unlock_irqrestore(&xhci->lock, flags); - kfree(command); + xhci_free_command(xhci, command); return -ENODEV; } @@ -4145,7 +4145,7 @@ int xhci_disable_slot(struct xhci_hcd *xhci, u32 slot_id) slot_id); if (ret) { spin_unlock_irqrestore(&xhci->lock, flags); - kfree(command); + xhci_free_command(xhci, command); return ret; } xhci_ring_cmd_db(xhci); -- 2.34.1

3 days, 8 hours

[PATCH] io_uring/rsrc: fix RLIMIT_MEMLOCK bypass via compound page accounting

by Yuhao Jiang

When multiple registered buffers share the same compound page, only the first buffer accounts for the memory via io_buffer_account_pin(). The subsequent buffers skip accounting since headpage_already_acct() returns true. When the first buffer is unregistered, the accounting is decremented, but the compound page remains pinned by the remaining buffers. This creates a state where pinned memory is not properly accounted against RLIMIT_MEMLOCK. On systems with HugeTLB pages pre-allocated, an unprivileged user can exploit this to pin memory beyond RLIMIT_MEMLOCK by cycling buffer registrations. The bypass amount is proportional to the number of available huge pages, potentially allowing gigabytes of memory to be pinned while the kernel accounting shows near-zero. Fix this by recalculating the actual pages to unaccount when unmapping a buffer. For regular pages, always unaccount. For compound pages, only unaccount if no other registered buffer references the same compound page. This ensures the accounting persists until the last buffer referencing the compound page is released. Reported-by: Yuhao Jiang <danisjiang(a)gmail.com> Fixes: 57bebf807e2a ("io_uring/rsrc: optimise registered huge pages") Cc: stable(a)vger.kernel.org Signed-off-by: Yuhao Jiang <danisjiang(a)gmail.com> --- io_uring/rsrc.c | 69 +++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 67 insertions(+), 2 deletions(-) diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c index a63474b331bf..dcf2340af5a2 100644 --- a/io_uring/rsrc.c +++ b/io_uring/rsrc.c @@ -139,15 +139,80 @@ static void io_free_imu(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu) kvfree(imu); } +/* + * Calculate pages to unaccount when unmapping a buffer. Regular pages are + * always counted. Compound pages are only counted if no other registered + * buffer references them, ensuring accounting persists until the last user. + */ +static unsigned long io_buffer_calc_unaccount(struct io_ring_ctx *ctx, + struct io_mapped_ubuf *imu) +{ + struct page *last_hpage = NULL; + unsigned long acct = 0; + unsigned int i; + + for (i = 0; i < imu->nr_bvecs; i++) { + struct page *page = imu->bvec[i].bv_page; + struct page *hpage; + unsigned int j; + + if (!PageCompound(page)) { + acct++; + continue; + } + + hpage = compound_head(page); + if (hpage == last_hpage) + continue; + last_hpage = hpage; + + /* Check if we already processed this hpage earlier in this buffer */ + for (j = 0; j < i; j++) { + if (PageCompound(imu->bvec[j].bv_page) && + compound_head(imu->bvec[j].bv_page) == hpage) + goto next_hpage; + } + + /* Only unaccount if no other buffer references this page */ + for (j = 0; j < ctx->buf_table.nr; j++) { + struct io_rsrc_node *node = ctx->buf_table.nodes[j]; + struct io_mapped_ubuf *other; + unsigned int k; + + if (!node) + continue; + other = node->buf; + if (other == imu) + continue; + + for (k = 0; k < other->nr_bvecs; k++) { + struct page *op = other->bvec[k].bv_page; + + if (!PageCompound(op)) + continue; + if (compound_head(op) == hpage) + goto next_hpage; + } + } + acct += page_size(hpage) >> PAGE_SHIFT; +next_hpage: + ; + } + return acct; +} + static void io_buffer_unmap(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu) { + unsigned long acct; + if (unlikely(refcount_read(&imu->refs) > 1)) { if (!refcount_dec_and_test(&imu->refs)) return; } - if (imu->acct_pages) - io_unaccount_mem(ctx->user, ctx->mm_account, imu->acct_pages); + acct = io_buffer_calc_unaccount(ctx, imu); + if (acct) + io_unaccount_mem(ctx->user, ctx->mm_account, acct); imu->release(imu->priv); io_free_imu(ctx, imu); } -- 2.34.1

3 days, 8 hours

[PATCH V2 5.4.y 0/2] Fix bad pmd due to race between change_prot_numa() and THP migration

by Harry Yoo

V1 -> V2: - Because `pmd_val` variable broke ppc builds due to its name, renamed it to `_pmd`. see [1]. [1] https://lore.kernel.org/stable/aS7lPZPYuChOTdXU@hyeyoo - Added David Hildenbrand's Acked-by [2], thanks a lot! [2] https://lore.kernel.org/linux-mm/ac8d7137-3819-4a75-9dd3-fb3d2259ebe4@kerne… # TL;DR previous discussion: https://lore.kernel.org/linux-mm/20250921232709.1608699-1-harry.yoo@oracle.… A "bad pmd" error occurs due to race condition between change_prot_numa() and THP migration. The mainline kernel does not have this bug as commit 670ddd8cdc fixes the race condition. 6.1.y, 5.15.y, 5.10.y, 5.4.y are affected by this bug. Fixing this in -stable kernels is tricky because pte_map_offset_lock() has different semantics in pre-6.5 and post-6.5 kernels. I am trying to backport the same mechanism we have in the mainline kernel. Since the code looks bit different due to different semantics of pte_map_offset_lock(), it'd be best to get this reviewed by MM folks. # Testing I verified that the bug described below is not reproduced anymore (on a downstream kernel) after applying this patch series. It used to trigger in few days of intensive numa balancing testing, but it survived 2 weeks with this applied. # Bug Description It was reported that a bad pmd is seen when automatic NUMA balancing is marking page table entries as prot_numa: [2437548.196018] mm/pgtable-generic.c:50: bad pmd 00000000af22fc02(dffffffe71fbfe02) [2437548.235022] Call Trace: [2437548.238234] <TASK> [2437548.241060] dump_stack_lvl+0x46/0x61 [2437548.245689] panic+0x106/0x2e5 [2437548.249497] pmd_clear_bad+0x3c/0x3c [2437548.253967] change_pmd_range.isra.0+0x34d/0x3a7 [2437548.259537] change_p4d_range+0x156/0x20e [2437548.264392] change_protection_range+0x116/0x1a9 [2437548.269976] change_prot_numa+0x15/0x37 [2437548.274774] task_numa_work+0x1b8/0x302 [2437548.279512] task_work_run+0x62/0x95 [2437548.283882] exit_to_user_mode_loop+0x1a4/0x1a9 [2437548.289277] exit_to_user_mode_prepare+0xf4/0xfc [2437548.294751] ? sysvec_apic_timer_interrupt+0x34/0x81 [2437548.300677] irqentry_exit_to_user_mode+0x5/0x25 [2437548.306153] asm_sysvec_apic_timer_interrupt+0x16/0x1b This is due to a race condition between change_prot_numa() and THP migration because the kernel doesn't check is_swap_pmd() and pmd_trans_huge() atomically: change_prot_numa() THP migration ====================================================================== - change_pmd_range() -> is_swap_pmd() returns false, meaning it's not a PMD migration entry. - do_huge_pmd_numa_page() -> migrate_misplaced_page() sets migration entries for the THP. - change_pmd_range() -> pmd_none_or_clear_bad_unless_trans_huge() -> pmd_none() and pmd_trans_huge() returns false - pmd_none_or_clear_bad_unless_trans_huge() -> pmd_bad() returns true for the migration entry! The upstream commit 670ddd8cdcbd ("mm/mprotect: delete pmd_none_or_clear_bad_unless_trans_huge()") closes this race condition by checking is_swap_pmd() and pmd_trans_huge() atomically. # Backporting note commit a79390f5d6a7 ("mm/mprotect: use long for page accountings and retval") is backported to return an error code (negative value) in change_pte_range(). Unlike the mainline, pte_offset_map_lock() does not check if the pmd entry is a migration entry or a hugepage; acquires PTL unconditionally instead of returning failure. Therefore, it is necessary to keep the !is_swap_pmd() && !pmd_trans_huge() && !pmd_devmap() checks in change_pmd_range() before acquiring the PTL. After acquiring the lock, open-code the semantics of pte_offset_map_lock() in the mainline kernel; change_pte_range() fails if the pmd value has changed. This requires adding pmd_old parameter (pmd_t value that is read before calling the function) to change_pte_range(). Hugh Dickins (1): mm/mprotect: delete pmd_none_or_clear_bad_unless_trans_huge() Peter Xu (1): mm/mprotect: use long for page accountings and retval include/linux/hugetlb.h | 4 +- include/linux/mm.h | 2 +- mm/hugetlb.c | 4 +- mm/mempolicy.c | 2 +- mm/mprotect.c | 124 +++++++++++++++++----------------------- 5 files changed, 60 insertions(+), 76 deletions(-) -- 2.43.0

3 days, 10 hours

[6.6-stable 0/2] mm: two fixes for bdi min_ratio and max_ratio knob

by Jingbo Xu

Have no idea why stable branch missed the "fix" tag. Jingbo Xu (2): mm: fix arithmetic for bdi min_ratio mm: fix arithmetic for max_prop_frac when setting max_ratio mm/page-writeback.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) -- 2.19.1.6.gb485710b

3 days, 10 hours

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror