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

[PATCH 02/14] drm/amd/display: Correct the reply value when AUX write incomplete

by Ray Wu

From: Wayne Lin <Wayne.Lin(a)amd.com> [Why] Now forcing aux->transfer to return 0 when incomplete AUX write is inappropriate. It should return bytes have been transferred. [How] aux->transfer is asked not to change original msg except reply field of drm_dp_aux_msg structure. Copy the msg->buffer when it's write request, and overwrite the first byte when sink reply 1 byte indicating partially written byte number. Then we can return the correct value without changing the original msg. Fixes: 6285f12bc54c ("drm/amd/display: Fix wrong handling for AUX_DEFER case") Cc: stable(a)vger.kernel.org Cc: Mario Limonciello <mario.limonciello(a)amd.com> Cc: Alex Deucher <alexander.deucher(a)amd.com> Reviewed-by: Ray Wu <ray.wu(a)amd.com> Signed-off-by: Wayne Lin <Wayne.Lin(a)amd.com> Signed-off-by: Ray Wu <ray.wu(a)amd.com> --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 3 ++- .../drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 10 ++++++++-- 2 files changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 8984e211dd1c..36c16030fca9 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -12853,7 +12853,8 @@ int amdgpu_dm_process_dmub_aux_transfer_sync( /* The reply is stored in the top nibble of the command. */ payload->reply[0] = (adev->dm.dmub_notify->aux_reply.command >> 4) & 0xF; - if (!payload->write && p_notify->aux_reply.length) + /*write req may receive a byte indicating partially written number as well*/ + if (p_notify->aux_reply.length) memcpy(payload->data, p_notify->aux_reply.data, p_notify->aux_reply.length); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index d19aea595722..0d7b72c75802 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -62,6 +62,7 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, enum aux_return_code_type operation_result; struct amdgpu_device *adev; struct ddc_service *ddc; + uint8_t copy[16]; if (WARN_ON(msg->size > 16)) return -E2BIG; @@ -77,6 +78,11 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, (msg->request & DP_AUX_I2C_WRITE_STATUS_UPDATE) != 0; payload.defer_delay = 0; + if (payload.write) { + memcpy(copy, msg->buffer, msg->size); + payload.data = copy; + } + result = dc_link_aux_transfer_raw(TO_DM_AUX(aux)->ddc_service, &payload, &operation_result); @@ -100,9 +106,9 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, */ if (payload.write && result >= 0) { if (result) { - /*one byte indicating partially written bytes. Force 0 to retry*/ + /*one byte indicating partially written bytes*/ drm_info(adev_to_drm(adev), "amdgpu: AUX partially written\n"); - result = 0; + result = payload.data[0]; } else if (!payload.reply[0]) /*I2C_ACK|AUX_ACK*/ result = msg->size; -- 2.43.0

2 months

1
0
0 0

[REGRESSION][BISECTED][STABLE] MT7925: mDNS and IPv6 broken in kernel 6.14.3 and above

by fossben＠pm.me

Hello all, After upgrading to 6.14.3 on my PC with a MT7925 chip, I noticed that I could no longer ping *.local addresses provided by Avahi. In addition, I also noticed that I was not able to get a DHCP IPv6 address from my router, no matter how many times I rebooted the router or reconnected with NetworkManager. Reverting to 6.14.2 fixes both mDNS and IPv6 addresses immediately. Going back to 6.14.3 immediately breaks mDNS again, but the IPv6 address will stay there for a while before disappearing later, possibly because the DHCP lease expired? I am not sure exactly when it stops working. I've done a kernel bisect between 6.14.2 and 6.14.3 and found the offending commit that causes mDNS to fail: commit 80007d3f92fd018d0a052a706400e976b36e3c87 Author: Ming Yen Hsieh <mingyen.hsieh(a)mediatek.com> Date: Tue Mar 4 16:08:50 2025 -0800 wifi: mt76: mt7925: integrate *mlo_sta_cmd and *sta_cmd commit cb1353ef34735ec1e5d9efa1fe966f05ff1dc1e1 upstream. Integrate *mlo_sta_cmd and *sta_cmd for the MLO firmware. Fixes: 86c051f2c418 ("wifi: mt76: mt7925: enabling MLO when the firmware supports it") drivers/net/wireless/mediatek/mt76/mt7925/mcu.c | 59 ++++------------------------------------------------------- 1 file changed, 4 insertions(+), 55 deletions(-) I do not know if this same commit is also causing the IPv6 issues as testing that requires quite a bit of time to reproduce. What I do know with certainty as of this moment is that it definitely breaks in kernel 6.14.3. I've attached my hardware info as well as dmesg logs from the last working kernel from the bisect and 6.14.4 which exhibits the issue. Please let me know if there's any other info you need. Thanks! Benjamin Xiao

2 months

3
9
0 0

[PATCH v2] usb: typec: tcpm: move tcpm_queue_vdm_unlocked to asynchronous work

by RD Babiera

A state check was previously added to tcpm_queue_vdm_unlocked to prevent a deadlock where the DisplayPort Alt Mode driver would be executing work and attempting to grab the tcpm_lock while the TCPM was holding the lock and attempting to unregister the altmode, blocking on the altmode driver's cancel_work_sync call. Because the state check isn't protected, there is a small window where the Alt Mode driver could determine that the TCPM is in a ready state and attempt to grab the lock while the TCPM grabs the lock and changes the TCPM state to one that causes the deadlock. The callstack is provided below: [110121.667392][ C7] Call trace: [110121.667396][ C7] __switch_to+0x174/0x338 [110121.667406][ C7] __schedule+0x608/0x9f0 [110121.667414][ C7] schedule+0x7c/0xe8 [110121.667423][ C7] kernfs_drain+0xb0/0x114 [110121.667431][ C7] __kernfs_remove+0x16c/0x20c [110121.667436][ C7] kernfs_remove_by_name_ns+0x74/0xe8 [110121.667442][ C7] sysfs_remove_group+0x84/0xe8 [110121.667450][ C7] sysfs_remove_groups+0x34/0x58 [110121.667458][ C7] device_remove_groups+0x10/0x20 [110121.667464][ C7] device_release_driver_internal+0x164/0x2e4 [110121.667475][ C7] device_release_driver+0x18/0x28 [110121.667484][ C7] bus_remove_device+0xec/0x118 [110121.667491][ C7] device_del+0x1e8/0x4ac [110121.667498][ C7] device_unregister+0x18/0x38 [110121.667504][ C7] typec_unregister_altmode+0x30/0x44 [110121.667515][ C7] tcpm_reset_port+0xac/0x370 [110121.667523][ C7] tcpm_snk_detach+0x84/0xb8 [110121.667529][ C7] run_state_machine+0x4c0/0x1b68 [110121.667536][ C7] tcpm_state_machine_work+0x94/0xe4 [110121.667544][ C7] kthread_worker_fn+0x10c/0x244 [110121.667552][ C7] kthread+0x104/0x1d4 [110121.667557][ C7] ret_from_fork+0x10/0x20 [110121.667689][ C7] Workqueue: events dp_altmode_work [110121.667697][ C7] Call trace: [110121.667701][ C7] __switch_to+0x174/0x338 [110121.667710][ C7] __schedule+0x608/0x9f0 [110121.667717][ C7] schedule+0x7c/0xe8 [110121.667725][ C7] schedule_preempt_disabled+0x24/0x40 [110121.667733][ C7] __mutex_lock+0x408/0xdac [110121.667741][ C7] __mutex_lock_slowpath+0x14/0x24 [110121.667748][ C7] mutex_lock+0x40/0xec [110121.667757][ C7] tcpm_altmode_enter+0x78/0xb4 [110121.667764][ C7] typec_altmode_enter+0xdc/0x10c [110121.667769][ C7] dp_altmode_work+0x68/0x164 [110121.667775][ C7] process_one_work+0x1e4/0x43c [110121.667783][ C7] worker_thread+0x25c/0x430 [110121.667789][ C7] kthread+0x104/0x1d4 [110121.667794][ C7] ret_from_fork+0x10/0x20 Change tcpm_queue_vdm_unlocked to queue for tcpm_queue_vdm_work, which can perform the state check while holding the TCPM lock while the Alt Mode lock is no longer held. This requires a new struct to hold the vdm data, altmode_vdm_event. Fixes: cdc9946ea637 ("usb: typec: tcpm: enforce ready state when queueing alt mode vdm") Cc: stable(a)vger.kernel.org Signed-off-by: RD Babiera <rdbabiera(a)google.com> Reviewed-by: Heikki Krogerus <heikki.krogerus(a)linux.intel.com> --- Changes from v1: * modified commit message to include call stack --- drivers/usb/typec/tcpm/tcpm.c | 91 +++++++++++++++++++++++++++-------- 1 file changed, 71 insertions(+), 20 deletions(-) diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c index 784fa23102f9..9b8d98328ddb 100644 --- a/drivers/usb/typec/tcpm/tcpm.c +++ b/drivers/usb/typec/tcpm/tcpm.c @@ -597,6 +597,15 @@ struct pd_rx_event { enum tcpm_transmit_type rx_sop_type; }; +struct altmode_vdm_event { + struct kthread_work work; + struct tcpm_port *port; + u32 header; + u32 *data; + int cnt; + enum tcpm_transmit_type tx_sop_type; +}; + static const char * const pd_rev[] = { [PD_REV10] = "rev1", [PD_REV20] = "rev2", @@ -1610,18 +1619,68 @@ static void tcpm_queue_vdm(struct tcpm_port *port, const u32 header, mod_vdm_delayed_work(port, 0); } -static void tcpm_queue_vdm_unlocked(struct tcpm_port *port, const u32 header, - const u32 *data, int cnt, enum tcpm_transmit_type tx_sop_type) +static void tcpm_queue_vdm_work(struct kthread_work *work) { - if (port->state != SRC_READY && port->state != SNK_READY && - port->state != SRC_VDM_IDENTITY_REQUEST) - return; + struct altmode_vdm_event *event = container_of(work, + struct altmode_vdm_event, + work); + struct tcpm_port *port = event->port; mutex_lock(&port->lock); - tcpm_queue_vdm(port, header, data, cnt, tx_sop_type); + if (port->state != SRC_READY && port->state != SNK_READY && + port->state != SRC_VDM_IDENTITY_REQUEST) { + tcpm_log_force(port, "dropping altmode_vdm_event"); + goto port_unlock; + } + + tcpm_queue_vdm(port, event->header, event->data, event->cnt, event->tx_sop_type); + +port_unlock: + kfree(event->data); + kfree(event); mutex_unlock(&port->lock); } +static int tcpm_queue_vdm_unlocked(struct tcpm_port *port, const u32 header, + const u32 *data, int cnt, enum tcpm_transmit_type tx_sop_type) +{ + struct altmode_vdm_event *event; + u32 *data_cpy; + int ret = -ENOMEM; + + event = kzalloc(sizeof(*event), GFP_KERNEL); + if (!event) + goto err_event; + + data_cpy = kcalloc(cnt, sizeof(u32), GFP_KERNEL); + if (!data_cpy) + goto err_data; + + kthread_init_work(&event->work, tcpm_queue_vdm_work); + event->port = port; + event->header = header; + memcpy(data_cpy, data, sizeof(u32) * cnt); + event->data = data_cpy; + event->cnt = cnt; + event->tx_sop_type = tx_sop_type; + + ret = kthread_queue_work(port->wq, &event->work); + if (!ret) { + ret = -EBUSY; + goto err_queue; + } + + return 0; + +err_queue: + kfree(data_cpy); +err_data: + kfree(event); +err_event: + tcpm_log_force(port, "failed to queue altmode vdm, err:%d", ret); + return ret; +} + static void svdm_consume_identity(struct tcpm_port *port, const u32 *p, int cnt) { u32 vdo = p[VDO_INDEX_IDH]; @@ -2832,8 +2891,7 @@ static int tcpm_altmode_enter(struct typec_altmode *altmode, u32 *vdo) header = VDO(altmode->svid, vdo ? 2 : 1, svdm_version, CMD_ENTER_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP); - return 0; + return tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP); } static int tcpm_altmode_exit(struct typec_altmode *altmode) @@ -2849,8 +2907,7 @@ static int tcpm_altmode_exit(struct typec_altmode *altmode) header = VDO(altmode->svid, 1, svdm_version, CMD_EXIT_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP); - return 0; + return tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP); } static int tcpm_altmode_vdm(struct typec_altmode *altmode, @@ -2858,9 +2915,7 @@ static int tcpm_altmode_vdm(struct typec_altmode *altmode, { struct tcpm_port *port = typec_altmode_get_drvdata(altmode); - tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP); - - return 0; + return tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP); } static const struct typec_altmode_ops tcpm_altmode_ops = { @@ -2884,8 +2939,7 @@ static int tcpm_cable_altmode_enter(struct typec_altmode *altmode, enum typec_pl header = VDO(altmode->svid, vdo ? 2 : 1, svdm_version, CMD_ENTER_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP_PRIME); - return 0; + return tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP_PRIME); } static int tcpm_cable_altmode_exit(struct typec_altmode *altmode, enum typec_plug_index sop) @@ -2901,8 +2955,7 @@ static int tcpm_cable_altmode_exit(struct typec_altmode *altmode, enum typec_plu header = VDO(altmode->svid, 1, svdm_version, CMD_EXIT_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP_PRIME); - return 0; + return tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP_PRIME); } static int tcpm_cable_altmode_vdm(struct typec_altmode *altmode, enum typec_plug_index sop, @@ -2910,9 +2963,7 @@ static int tcpm_cable_altmode_vdm(struct typec_altmode *altmode, enum typec_plug { struct tcpm_port *port = typec_altmode_get_drvdata(altmode); - tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP_PRIME); - - return 0; + return tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP_PRIME); } static const struct typec_cable_ops tcpm_cable_ops = { base-commit: 588d032e9e566997db3213dee145dbe3bda297b6 -- 2.49.0.967.g6a0df3ecc3-goog

2 months

2
1
0 0

[PATCH v2 0/2] Map MAP_STACK to VM_NOHUGEPAGE only if THP is enabled

by Ignacio Moreno Gonzalez via B4 Relay

... and make setting MADV_NOHUGEPAGE with madvise() into a no-op if THP is not enabled. I discovered this issue when trying to use the tool CRIU to checkpoint and restore a container. Our running kernel is compiled without CONFIG_TRANSPARENT_HUGETABLES. CRIU parses the output of /proc/<pid>/smaps and saves the "nh" flag. When trying to restore the container, CRIU fails to restore the "nh" mappings, since madvise() MADV_NOHUGEPAGE always returns an error because CONFIG_TRANSPARENT_HUGETABLES is not defined. These patches: - Avoid mapping MAP_STACK to VM_NOHUGEPAGE if !THP - Avoid returning an error when calling madvise() with MADV_NOHUGEPAGE if !THP Signed-off-by: Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez(a)kuka.com> --- Changes in v2: - [Patch 1/2] Use '#ifdef' instead of '#if defined(...)' - [Patch 1/2] Add 'Fixes: c4608d1bf7c6...' - Create [Patch 2/2] - Link to v1: https://lore.kernel.org/r/20250502-map-map_stack-to-vm_nohugepage-only-if-t… --- Ignacio Moreno Gonzalez (2): mm: mmap: map MAP_STACK to VM_NOHUGEPAGE only if THP is enabled mm: madvise: no-op for MADV_NOHUGEPAGE if THP is disabled include/linux/huge_mm.h | 6 ++++++ include/linux/mman.h | 2 ++ 2 files changed, 8 insertions(+) --- base-commit: fc96b232f8e7c0a6c282f47726b2ff6a5fb341d2 change-id: 20250428-map-map_stack-to-vm_nohugepage-only-if-thp-is-enabled-ce40a1de095d Best regards, -- Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez(a)kuka.com>

2 months

5
6
0 0

[PATCH v1] usb: typec: tcpm: move tcpm_queue_vdm_unlocked to asynchronous work

by RD Babiera

A state check was previously added to tcpm_queue_vdm_unlocked to prevent a deadlock where the DisplayPort Alt Mode driver would be executing work and attempting to grab the tcpm_lock while the TCPM was holding the lock and attempting to unregister the altmode, blocking on the altmode driver's cancel_work_sync call. Because the state check isn't protected, there is a small window where the Alt Mode driver could determine that the TCPM is in a ready state and attempt to grab the lock while the TCPM grabs the lock and changes the TCPM state to one that causes the deadlock. Change tcpm_queue_vdm_unlocked to queue for tcpm_queue_vdm_work, which can perform the state check while holding the TCPM lock while the Alt Mode lock is no longer held. This requires a new struct to hold the vdm data, altmode_vdm_event. Fixes: cdc9946ea637 ("usb: typec: tcpm: enforce ready state when queueing alt mode vdm") Cc: stable(a)vger.kernel.org Signed-off-by: RD Babiera <rdbabiera(a)google.com> --- drivers/usb/typec/tcpm/tcpm.c | 91 +++++++++++++++++++++++++++-------- 1 file changed, 71 insertions(+), 20 deletions(-) diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c index 784fa23102f9..9b8d98328ddb 100644 --- a/drivers/usb/typec/tcpm/tcpm.c +++ b/drivers/usb/typec/tcpm/tcpm.c @@ -597,6 +597,15 @@ struct pd_rx_event { enum tcpm_transmit_type rx_sop_type; }; +struct altmode_vdm_event { + struct kthread_work work; + struct tcpm_port *port; + u32 header; + u32 *data; + int cnt; + enum tcpm_transmit_type tx_sop_type; +}; + static const char * const pd_rev[] = { [PD_REV10] = "rev1", [PD_REV20] = "rev2", @@ -1610,18 +1619,68 @@ static void tcpm_queue_vdm(struct tcpm_port *port, const u32 header, mod_vdm_delayed_work(port, 0); } -static void tcpm_queue_vdm_unlocked(struct tcpm_port *port, const u32 header, - const u32 *data, int cnt, enum tcpm_transmit_type tx_sop_type) +static void tcpm_queue_vdm_work(struct kthread_work *work) { - if (port->state != SRC_READY && port->state != SNK_READY && - port->state != SRC_VDM_IDENTITY_REQUEST) - return; + struct altmode_vdm_event *event = container_of(work, + struct altmode_vdm_event, + work); + struct tcpm_port *port = event->port; mutex_lock(&port->lock); - tcpm_queue_vdm(port, header, data, cnt, tx_sop_type); + if (port->state != SRC_READY && port->state != SNK_READY && + port->state != SRC_VDM_IDENTITY_REQUEST) { + tcpm_log_force(port, "dropping altmode_vdm_event"); + goto port_unlock; + } + + tcpm_queue_vdm(port, event->header, event->data, event->cnt, event->tx_sop_type); + +port_unlock: + kfree(event->data); + kfree(event); mutex_unlock(&port->lock); } +static int tcpm_queue_vdm_unlocked(struct tcpm_port *port, const u32 header, + const u32 *data, int cnt, enum tcpm_transmit_type tx_sop_type) +{ + struct altmode_vdm_event *event; + u32 *data_cpy; + int ret = -ENOMEM; + + event = kzalloc(sizeof(*event), GFP_KERNEL); + if (!event) + goto err_event; + + data_cpy = kcalloc(cnt, sizeof(u32), GFP_KERNEL); + if (!data_cpy) + goto err_data; + + kthread_init_work(&event->work, tcpm_queue_vdm_work); + event->port = port; + event->header = header; + memcpy(data_cpy, data, sizeof(u32) * cnt); + event->data = data_cpy; + event->cnt = cnt; + event->tx_sop_type = tx_sop_type; + + ret = kthread_queue_work(port->wq, &event->work); + if (!ret) { + ret = -EBUSY; + goto err_queue; + } + + return 0; + +err_queue: + kfree(data_cpy); +err_data: + kfree(event); +err_event: + tcpm_log_force(port, "failed to queue altmode vdm, err:%d", ret); + return ret; +} + static void svdm_consume_identity(struct tcpm_port *port, const u32 *p, int cnt) { u32 vdo = p[VDO_INDEX_IDH]; @@ -2832,8 +2891,7 @@ static int tcpm_altmode_enter(struct typec_altmode *altmode, u32 *vdo) header = VDO(altmode->svid, vdo ? 2 : 1, svdm_version, CMD_ENTER_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP); - return 0; + return tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP); } static int tcpm_altmode_exit(struct typec_altmode *altmode) @@ -2849,8 +2907,7 @@ static int tcpm_altmode_exit(struct typec_altmode *altmode) header = VDO(altmode->svid, 1, svdm_version, CMD_EXIT_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP); - return 0; + return tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP); } static int tcpm_altmode_vdm(struct typec_altmode *altmode, @@ -2858,9 +2915,7 @@ static int tcpm_altmode_vdm(struct typec_altmode *altmode, { struct tcpm_port *port = typec_altmode_get_drvdata(altmode); - tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP); - - return 0; + return tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP); } static const struct typec_altmode_ops tcpm_altmode_ops = { @@ -2884,8 +2939,7 @@ static int tcpm_cable_altmode_enter(struct typec_altmode *altmode, enum typec_pl header = VDO(altmode->svid, vdo ? 2 : 1, svdm_version, CMD_ENTER_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP_PRIME); - return 0; + return tcpm_queue_vdm_unlocked(port, header, vdo, vdo ? 1 : 0, TCPC_TX_SOP_PRIME); } static int tcpm_cable_altmode_exit(struct typec_altmode *altmode, enum typec_plug_index sop) @@ -2901,8 +2955,7 @@ static int tcpm_cable_altmode_exit(struct typec_altmode *altmode, enum typec_plu header = VDO(altmode->svid, 1, svdm_version, CMD_EXIT_MODE); header |= VDO_OPOS(altmode->mode); - tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP_PRIME); - return 0; + return tcpm_queue_vdm_unlocked(port, header, NULL, 0, TCPC_TX_SOP_PRIME); } static int tcpm_cable_altmode_vdm(struct typec_altmode *altmode, enum typec_plug_index sop, @@ -2910,9 +2963,7 @@ static int tcpm_cable_altmode_vdm(struct typec_altmode *altmode, enum typec_plug { struct tcpm_port *port = typec_altmode_get_drvdata(altmode); - tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP_PRIME); - - return 0; + return tcpm_queue_vdm_unlocked(port, header, data, count - 1, TCPC_TX_SOP_PRIME); } static const struct typec_cable_ops tcpm_cable_ops = { base-commit: 615dca38c2eae55aff80050275931c87a812b48c -- 2.49.0.967.g6a0df3ecc3-goog

2 months

3
2
0 0

[PATCH v1 0/7] ublk: Backport to 6.14-stable: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd

by Jared Holzman

This patchset backports a series of ublk fixes from upstream to 6.14-stable. Patch 7 fixes the race that can cause kernel panic when ublk server daemon is exiting. It depends on patches 1-6 which simplifies & improves IO canceling when ublk server daemon is exiting as described here: https://lore.kernel.org/linux-block/20250416035444.99569-1-ming.lei@redhat.… Ming Lei (5): ublk: add helper of ublk_need_map_io() ublk: move device reset into ublk_ch_release() ublk: remove __ublk_quiesce_dev() ublk: simplify aborting ublk request ublk: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd Uday Shankar (2): ublk: properly serialize all FETCH_REQs ublk: improve detection and handling of ublk server exit drivers/block/ublk_drv.c | 550 +++++++++++++++++++++------------------ 1 file changed, 291 insertions(+), 259 deletions(-) -- 2.43.0

2 months

1
1
0 0

[PATCH v1 7/7] ublk: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd

by Jared Holzman

From: Ming Lei <ming.lei(a)redhat.com> ublk_cancel_cmd() calls io_uring_cmd_done() to complete uring_cmd, but we may have scheduled task work via io_uring_cmd_complete_in_task() for dispatching request, then kernel crash can be triggered. Fix it by not trying to canceling the command if ublk block request is started. Fixes: 216c8f5ef0f2 ("ublk: replace monitor with cancelable uring_cmd") Reported-by: Jared Holzman <jholzman(a)nvidia.com> Tested-by: Jared Holzman <jholzman(a)nvidia.com> Closes: https://lore.kernel.org/linux-block/d2179120-171b-47ba-b664-23242981ef19@nv… Signed-off-by: Ming Lei <ming.lei(a)redhat.com> Link: https://lore.kernel.org/r/20250425013742.1079549-3-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe(a)kernel.dk> --- drivers/block/ublk_drv.c | 27 +++++++++++++++++++++------ 1 file changed, 21 insertions(+), 6 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 6000147ac2a5..348c4feb7a2d 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -1655,14 +1655,31 @@ static void ublk_start_cancel(struct ublk_queue *ubq) ublk_put_disk(disk); } -static void ublk_cancel_cmd(struct ublk_queue *ubq, struct ublk_io *io, +static void ublk_cancel_cmd(struct ublk_queue *ubq, unsigned tag, unsigned int issue_flags) { + struct ublk_io *io = &ubq->ios[tag]; + struct ublk_device *ub = ubq->dev; + struct request *req; bool done; if (!(io->flags & UBLK_IO_FLAG_ACTIVE)) return; + /* + * Don't try to cancel this command if the request is started for + * avoiding race between io_uring_cmd_done() and + * io_uring_cmd_complete_in_task(). + * + * Either the started request will be aborted via __ublk_abort_rq(), + * then this uring_cmd is canceled next time, or it will be done in + * task work function ublk_dispatch_req() because io_uring guarantees + * that ublk_dispatch_req() is always called + */ + req = blk_mq_tag_to_rq(ub->tag_set.tags[ubq->q_id], tag); + if (req && blk_mq_request_started(req)) + return; + spin_lock(&ubq->cancel_lock); done = !!(io->flags & UBLK_IO_FLAG_CANCELED); if (!done) @@ -1694,7 +1711,6 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd, struct ublk_uring_cmd_pdu *pdu = ublk_get_uring_cmd_pdu(cmd); struct ublk_queue *ubq = pdu->ubq; struct task_struct *task; - struct ublk_io *io; if (WARN_ON_ONCE(!ubq)) return; @@ -1709,9 +1725,8 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd, if (!ubq->canceling) ublk_start_cancel(ubq); - io = &ubq->ios[pdu->tag]; - WARN_ON_ONCE(io->cmd != cmd); - ublk_cancel_cmd(ubq, io, issue_flags); + WARN_ON_ONCE(ubq->ios[pdu->tag].cmd != cmd); + ublk_cancel_cmd(ubq, pdu->tag, issue_flags); } static inline bool ublk_queue_ready(struct ublk_queue *ubq) @@ -1724,7 +1739,7 @@ static void ublk_cancel_queue(struct ublk_queue *ubq) int i; for (i = 0; i < ubq->q_depth; i++) - ublk_cancel_cmd(ubq, &ubq->ios[i], IO_URING_F_UNLOCKED); + ublk_cancel_cmd(ubq, i, IO_URING_F_UNLOCKED); } /* Cancel all pending commands, must be called after del_gendisk() returns */ -- 2.43.0

2 months

1
0
0 0

[PATCH v1 6/7] ublk: simplify aborting ublk request

by Jared Holzman

From: Ming Lei <ming.lei(a)redhat.com> Now ublk_abort_queue() is moved to ublk char device release handler, meantime our request queue is "quiesced" because either ->canceling was set from uring_cmd cancel function or all IOs are inflight and can't be completed by ublk server, things becomes easy much: - all uring_cmd are done, so we needn't to mark io as UBLK_IO_FLAG_ABORTED for handling completion from uring_cmd - ublk char device is closed, no one can hold IO request reference any more, so we can simply complete this request or requeue it for ublk_nosrv_should_reissue_outstanding. Reviewed-by: Uday Shankar <ushankar(a)purestorage.com> Signed-off-by: Ming Lei <ming.lei(a)redhat.com> Link: https://lore.kernel.org/r/20250416035444.99569-8-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe(a)kernel.dk> --- drivers/block/ublk_drv.c | 82 ++++++++++------------------------------ 1 file changed, 20 insertions(+), 62 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index c3f576a9dbf2..6000147ac2a5 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -115,15 +115,6 @@ struct ublk_uring_cmd_pdu { */ #define UBLK_IO_FLAG_OWNED_BY_SRV 0x02 -/* - * IO command is aborted, so this flag is set in case of - * !UBLK_IO_FLAG_ACTIVE. - * - * After this flag is observed, any pending or new incoming request - * associated with this io command will be failed immediately - */ -#define UBLK_IO_FLAG_ABORTED 0x04 - /* * UBLK_IO_FLAG_NEED_GET_DATA is set because IO command requires * get data buffer address from ublksrv. @@ -1054,12 +1045,6 @@ static inline void __ublk_complete_rq(struct request *req) unsigned int unmapped_bytes; blk_status_t res = BLK_STS_OK; - /* called from ublk_abort_queue() code path */ - if (io->flags & UBLK_IO_FLAG_ABORTED) { - res = BLK_STS_IOERR; - goto exit; - } - /* failed read IO if nothing is read */ if (!io->res && req_op(req) == REQ_OP_READ) io->res = -EIO; @@ -1109,47 +1094,6 @@ static void ublk_complete_rq(struct kref *ref) __ublk_complete_rq(req); } -static void ublk_do_fail_rq(struct request *req) -{ - struct ublk_queue *ubq = req->mq_hctx->driver_data; - - if (ublk_nosrv_should_reissue_outstanding(ubq->dev)) - blk_mq_requeue_request(req, false); - else - __ublk_complete_rq(req); -} - -static void ublk_fail_rq_fn(struct kref *ref) -{ - struct ublk_rq_data *data = container_of(ref, struct ublk_rq_data, - ref); - struct request *req = blk_mq_rq_from_pdu(data); - - ublk_do_fail_rq(req); -} - -/* - * Since ublk_rq_task_work_cb always fails requests immediately during - * exiting, __ublk_fail_req() is only called from abort context during - * exiting. So lock is unnecessary. - * - * Also aborting may not be started yet, keep in mind that one failed - * request may be issued by block layer again. - */ -static void __ublk_fail_req(struct ublk_queue *ubq, struct ublk_io *io, - struct request *req) -{ - WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_ACTIVE); - - if (ublk_need_req_ref(ubq)) { - struct ublk_rq_data *data = blk_mq_rq_to_pdu(req); - - kref_put(&data->ref, ublk_fail_rq_fn); - } else { - ublk_do_fail_rq(req); - } -} - static void ubq_complete_io_cmd(struct ublk_io *io, int res, unsigned issue_flags) { @@ -1639,10 +1583,26 @@ static void ublk_commit_completion(struct ublk_device *ub, ublk_put_req_ref(ubq, req); } +static void __ublk_fail_req(struct ublk_queue *ubq, struct ublk_io *io, + struct request *req) +{ + WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_ACTIVE); + + if (ublk_nosrv_should_reissue_outstanding(ubq->dev)) + blk_mq_requeue_request(req, false); + else { + io->res = -EIO; + __ublk_complete_rq(req); + } +} + /* - * Called from ubq_daemon context via cancel fn, meantime quiesce ublk - * blk-mq queue, so we are called exclusively with blk-mq and ubq_daemon - * context, so everything is serialized. + * Called from ublk char device release handler, when any uring_cmd is + * done, meantime request queue is "quiesced" since all inflight requests + * can't be completed because ublk server is dead. + * + * So no one can hold our request IO reference any more, simply ignore the + * reference, and complete the request immediately */ static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq) { @@ -1659,10 +1619,8 @@ static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq) * will do it */ rq = blk_mq_tag_to_rq(ub->tag_set.tags[ubq->q_id], i); - if (rq && blk_mq_request_started(rq)) { - io->flags |= UBLK_IO_FLAG_ABORTED; + if (rq && blk_mq_request_started(rq)) __ublk_fail_req(ubq, io, rq); - } } } } -- 2.43.0

2 months

1
0
0 0

[PATCH v1 5/7] ublk: remove __ublk_quiesce_dev()

by Jared Holzman

From: Ming Lei <ming.lei(a)redhat.com> Remove __ublk_quiesce_dev() and open code for updating device state as QUIESCED. We needn't to drain inflight requests in __ublk_quiesce_dev() any more, because all inflight requests are aborted in ublk char device release handler. Also we needn't to set ->canceling in __ublk_quiesce_dev() any more because it is done unconditionally now in ublk_ch_release(). Reviewed-by: Uday Shankar <ushankar(a)purestorage.com> Signed-off-by: Ming Lei <ming.lei(a)redhat.com> Link: https://lore.kernel.org/r/20250416035444.99569-7-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe(a)kernel.dk> --- drivers/block/ublk_drv.c | 19 ++----------------- 1 file changed, 2 insertions(+), 17 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 652742db0396..c3f576a9dbf2 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -205,7 +205,6 @@ struct ublk_params_header { static void ublk_stop_dev_unlocked(struct ublk_device *ub); static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq); -static void __ublk_quiesce_dev(struct ublk_device *ub); static inline unsigned int ublk_req_build_flags(struct request *req); static inline struct ublksrv_io_desc *ublk_get_iod(struct ublk_queue *ubq, @@ -1558,7 +1557,8 @@ static int ublk_ch_release(struct inode *inode, struct file *filp) ublk_stop_dev_unlocked(ub); } else { if (ublk_nosrv_dev_should_queue_io(ub)) { - __ublk_quiesce_dev(ub); + /* ->canceling is set and all requests are aborted */ + ub->dev_info.state = UBLK_S_DEV_QUIESCED; } else { ub->dev_info.state = UBLK_S_DEV_FAIL_IO; for (i = 0; i < ub->dev_info.nr_hw_queues; i++) @@ -1804,21 +1804,6 @@ static void ublk_wait_tagset_rqs_idle(struct ublk_device *ub) } } -static void __ublk_quiesce_dev(struct ublk_device *ub) -{ - int i; - - pr_devel("%s: quiesce ub: dev_id %d state %s\n", - __func__, ub->dev_info.dev_id, - ub->dev_info.state == UBLK_S_DEV_LIVE ? - "LIVE" : "QUIESCED"); - /* mark every queue as canceling */ - for (i = 0; i < ub->dev_info.nr_hw_queues; i++) - ublk_get_queue(ub, i)->canceling = true; - ublk_wait_tagset_rqs_idle(ub); - ub->dev_info.state = UBLK_S_DEV_QUIESCED; -} - static void ublk_force_abort_dev(struct ublk_device *ub) { int i; -- 2.43.0

2 months

1
0
0 0

[PATCH v1 4/7] ublk: improve detection and handling of ublk server exit

by Jared Holzman

From: Uday Shankar <ushankar(a)purestorage.com> There are currently two ways in which ublk server exit is detected by ublk_drv: 1. uring_cmd cancellation. If there are any outstanding uring_cmds which have not been completed to the ublk server when it exits, io_uring calls the uring_cmd callback with a special cancellation flag as the issuing task is exiting. 2. I/O timeout. This is needed in addition to the above to handle the "saturated queue" case, when all I/Os for a given queue are in the ublk server, and therefore there are no outstanding uring_cmds to cancel when the ublk server exits. There are a couple of issues with this approach: - It is complex and inelegant to have two methods to detect the same condition - The second method detects ublk server exit only after a long delay (~30s, the default timeout assigned by the block layer). This delays the nosrv behavior from kicking in and potential subsequent recovery of the device. The second issue is brought to light with the new test_generic_06 which will be added in following patch. It fails before this fix: selftests: ublk: test_generic_06.sh dev id is 0 dd: error writing '/dev/ublkb0': Input/output error 1+0 records in 0+0 records out 0 bytes copied, 30.0611 s, 0.0 kB/s DEAD dd took 31 seconds to exit (>= 5s tolerance)! generic_06 : [FAIL] Fix this by instead detecting and handling ublk server exit in the character file release callback. This has several advantages: - This one place can handle both saturated and unsaturated queues. Thus, it replaces both preexisting methods of detecting ublk server exit. - It runs quickly on ublk server exit - there is no 30s delay. - It starts the process of removing task references in ublk_drv. This is needed if we want to relax restrictions in the driver like letting only one thread serve each queue There is also the disadvantage that the character file release callback can also be triggered by intentional close of the file, which is a significant behavior change. Preexisting ublk servers (libublksrv) are dependent on the ability to open/close the file multiple times. To address this, only transition to a nosrv state if the file is released while the ublk device is live. This allows for programs to open/close the file multiple times during setup. It is still a behavior change if a ublk server decides to close/reopen the file while the device is LIVE (i.e. while it is responsible for serving I/O), but that would be highly unusual. This behavior is in line with what is done by FUSE, which is very similar to ublk in that a userspace daemon is providing services traditionally provided by the kernel. With this change in, the new test (and all other selftests, and all ublksrv tests) pass: selftests: ublk: test_generic_06.sh dev id is 0 dd: error writing '/dev/ublkb0': Input/output error 1+0 records in 0+0 records out 0 bytes copied, 0.0376731 s, 0.0 kB/s DEAD generic_04 : [PASS] Signed-off-by: Uday Shankar <ushankar(a)purestorage.com> Signed-off-by: Ming Lei <ming.lei(a)redhat.com> Link: https://lore.kernel.org/r/20250416035444.99569-6-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe(a)kernel.dk> --- drivers/block/ublk_drv.c | 223 ++++++++++++++++++++++----------------- 1 file changed, 124 insertions(+), 99 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index c619df880c72..652742db0396 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -194,8 +194,6 @@ struct ublk_device { struct completion completion; unsigned int nr_queues_ready; unsigned int nr_privileged_daemon; - - struct work_struct nosrv_work; }; /* header of ublk_params */ @@ -204,7 +202,10 @@ struct ublk_params_header { __u32 types; }; -static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq); + +static void ublk_stop_dev_unlocked(struct ublk_device *ub); +static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq); +static void __ublk_quiesce_dev(struct ublk_device *ub); static inline unsigned int ublk_req_build_flags(struct request *req); static inline struct ublksrv_io_desc *ublk_get_iod(struct ublk_queue *ubq, @@ -1306,8 +1307,6 @@ static void ublk_queue_cmd_list(struct ublk_queue *ubq, struct rq_list *l) static enum blk_eh_timer_return ublk_timeout(struct request *rq) { struct ublk_queue *ubq = rq->mq_hctx->driver_data; - unsigned int nr_inflight = 0; - int i; if (ubq->flags & UBLK_F_UNPRIVILEGED_DEV) { if (!ubq->timeout) { @@ -1318,26 +1317,6 @@ static enum blk_eh_timer_return ublk_timeout(struct request *rq) return BLK_EH_DONE; } - if (!ubq_daemon_is_dying(ubq)) - return BLK_EH_RESET_TIMER; - - for (i = 0; i < ubq->q_depth; i++) { - struct ublk_io *io = &ubq->ios[i]; - - if (!(io->flags & UBLK_IO_FLAG_ACTIVE)) - nr_inflight++; - } - - /* cancelable uring_cmd can't help us if all commands are in-flight */ - if (nr_inflight == ubq->q_depth) { - struct ublk_device *ub = ubq->dev; - - if (ublk_abort_requests(ub, ubq)) { - schedule_work(&ub->nosrv_work); - } - return BLK_EH_DONE; - } - return BLK_EH_RESET_TIMER; } @@ -1495,13 +1474,105 @@ static void ublk_reset_ch_dev(struct ublk_device *ub) ub->nr_privileged_daemon = 0; } +static struct gendisk *ublk_get_disk(struct ublk_device *ub) +{ + struct gendisk *disk; + + spin_lock(&ub->lock); + disk = ub->ub_disk; + if (disk) + get_device(disk_to_dev(disk)); + spin_unlock(&ub->lock); + + return disk; +} + +static void ublk_put_disk(struct gendisk *disk) +{ + if (disk) + put_device(disk_to_dev(disk)); +} + static int ublk_ch_release(struct inode *inode, struct file *filp) { struct ublk_device *ub = filp->private_data; + struct gendisk *disk; + int i; + + /* + * disk isn't attached yet, either device isn't live, or it has + * been removed already, so we needn't to do anything + */ + disk = ublk_get_disk(ub); + if (!disk) + goto out; + + /* + * All uring_cmd are done now, so abort any request outstanding to + * the ublk server + * + * This can be done in lockless way because ublk server has been + * gone + * + * More importantly, we have to provide forward progress guarantee + * without holding ub->mutex, otherwise control task grabbing + * ub->mutex triggers deadlock + * + * All requests may be inflight, so ->canceling may not be set, set + * it now. + */ + for (i = 0; i < ub->dev_info.nr_hw_queues; i++) { + struct ublk_queue *ubq = ublk_get_queue(ub, i); + + ubq->canceling = true; + ublk_abort_queue(ub, ubq); + } + blk_mq_kick_requeue_list(disk->queue); + + /* + * All infligh requests have been completed or requeued and any new + * request will be failed or requeued via `->canceling` now, so it is + * fine to grab ub->mutex now. + */ + mutex_lock(&ub->mutex); + + /* double check after grabbing lock */ + if (!ub->ub_disk) + goto unlock; + + /* + * Transition the device to the nosrv state. What exactly this + * means depends on the recovery flags + */ + blk_mq_quiesce_queue(disk->queue); + if (ublk_nosrv_should_stop_dev(ub)) { + /* + * Allow any pending/future I/O to pass through quickly + * with an error. This is needed because del_gendisk + * waits for all pending I/O to complete + */ + for (i = 0; i < ub->dev_info.nr_hw_queues; i++) + ublk_get_queue(ub, i)->force_abort = true; + blk_mq_unquiesce_queue(disk->queue); + + ublk_stop_dev_unlocked(ub); + } else { + if (ublk_nosrv_dev_should_queue_io(ub)) { + __ublk_quiesce_dev(ub); + } else { + ub->dev_info.state = UBLK_S_DEV_FAIL_IO; + for (i = 0; i < ub->dev_info.nr_hw_queues; i++) + ublk_get_queue(ub, i)->fail_io = true; + } + blk_mq_unquiesce_queue(disk->queue); + } +unlock: + mutex_unlock(&ub->mutex); + ublk_put_disk(disk); /* all uring_cmd has been done now, reset device & ubq */ ublk_reset_ch_dev(ub); - +out: clear_bit(UB_STATE_OPEN, &ub->state); return 0; } @@ -1597,37 +1668,22 @@ static void ublk_abort_queue(struct ublk_device *ub, struct ublk_queue *ubq) } /* Must be called when queue is frozen */ -static bool ublk_mark_queue_canceling(struct ublk_queue *ubq) +static void ublk_mark_queue_canceling(struct ublk_queue *ubq) { - bool canceled; - spin_lock(&ubq->cancel_lock); - canceled = ubq->canceling; - if (!canceled) + if (!ubq->canceling) ubq->canceling = true; spin_unlock(&ubq->cancel_lock); - - return canceled; } -static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq) +static void ublk_start_cancel(struct ublk_queue *ubq) { - bool was_canceled = ubq->canceling; - struct gendisk *disk; - - if (was_canceled) - return false; - - spin_lock(&ub->lock); - disk = ub->ub_disk; - if (disk) - get_device(disk_to_dev(disk)); - spin_unlock(&ub->lock); + struct ublk_device *ub = ubq->dev; + struct gendisk *disk = ublk_get_disk(ub); /* Our disk has been dead */ if (!disk) - return false; - + return; /* * Now we are serialized with ublk_queue_rq() * @@ -1636,15 +1692,9 @@ static bool ublk_abort_requests(struct ublk_device *ub, struct ublk_queue *ubq) * touch completed uring_cmd */ blk_mq_quiesce_queue(disk->queue); - was_canceled = ublk_mark_queue_canceling(ubq); - if (!was_canceled) { - /* abort queue is for making forward progress */ - ublk_abort_queue(ub, ubq); - } + ublk_mark_queue_canceling(ubq); blk_mq_unquiesce_queue(disk->queue); - put_device(disk_to_dev(disk)); - - return !was_canceled; + ublk_put_disk(disk); } static void ublk_cancel_cmd(struct ublk_queue *ubq, struct ublk_io *io, @@ -1668,6 +1718,17 @@ static void ublk_cancel_cmd(struct ublk_queue *ubq, struct ublk_io *io, /* * The ublk char device won't be closed when calling cancel fn, so both * ublk device and queue are guaranteed to be live + * + * Two-stage cancel: + * + * - make every active uring_cmd done in ->cancel_fn() + * + * - aborting inflight ublk IO requests in ublk char device release handler, + * which depends on 1st stage because device can only be closed iff all + * uring_cmd are done + * + * Do _not_ try to acquire ub->mutex before all inflight requests are + * aborted, otherwise deadlock may be caused. */ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd, unsigned int issue_flags) @@ -1675,8 +1736,6 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd, struct ublk_uring_cmd_pdu *pdu = ublk_get_uring_cmd_pdu(cmd); struct ublk_queue *ubq = pdu->ubq; struct task_struct *task; - struct ublk_device *ub; - bool need_schedule; struct ublk_io *io; if (WARN_ON_ONCE(!ubq)) @@ -1689,16 +1748,12 @@ static void ublk_uring_cmd_cancel_fn(struct io_uring_cmd *cmd, if (WARN_ON_ONCE(task && task != ubq->ubq_daemon)) return; - ub = ubq->dev; - need_schedule = ublk_abort_requests(ub, ubq); + if (!ubq->canceling) + ublk_start_cancel(ubq); io = &ubq->ios[pdu->tag]; WARN_ON_ONCE(io->cmd != cmd); ublk_cancel_cmd(ubq, io, issue_flags); - - if (need_schedule) { - schedule_work(&ub->nosrv_work); - } } static inline bool ublk_queue_ready(struct ublk_queue *ubq) @@ -1757,13 +1812,11 @@ static void __ublk_quiesce_dev(struct ublk_device *ub) __func__, ub->dev_info.dev_id, ub->dev_info.state == UBLK_S_DEV_LIVE ? "LIVE" : "QUIESCED"); - blk_mq_quiesce_queue(ub->ub_disk->queue); /* mark every queue as canceling */ for (i = 0; i < ub->dev_info.nr_hw_queues; i++) ublk_get_queue(ub, i)->canceling = true; ublk_wait_tagset_rqs_idle(ub); ub->dev_info.state = UBLK_S_DEV_QUIESCED; - blk_mq_unquiesce_queue(ub->ub_disk->queue); } static void ublk_force_abort_dev(struct ublk_device *ub) @@ -1800,50 +1853,25 @@ static struct gendisk *ublk_detach_disk(struct ublk_device *ub) return disk; } -static void ublk_stop_dev(struct ublk_device *ub) +static void ublk_stop_dev_unlocked(struct ublk_device *ub) + __must_hold(&ub->mutex) { struct gendisk *disk; - mutex_lock(&ub->mutex); if (ub->dev_info.state == UBLK_S_DEV_DEAD) - goto unlock; + return; + if (ublk_nosrv_dev_should_queue_io(ub)) ublk_force_abort_dev(ub); del_gendisk(ub->ub_disk); disk = ublk_detach_disk(ub); put_disk(disk); - unlock: - mutex_unlock(&ub->mutex); - ublk_cancel_dev(ub); } -static void ublk_nosrv_work(struct work_struct *work) +static void ublk_stop_dev(struct ublk_device *ub) { - struct ublk_device *ub = - container_of(work, struct ublk_device, nosrv_work); - int i; - - if (ublk_nosrv_should_stop_dev(ub)) { - ublk_stop_dev(ub); - return; - } - mutex_lock(&ub->mutex); - if (ub->dev_info.state != UBLK_S_DEV_LIVE) - goto unlock; - - if (ublk_nosrv_dev_should_queue_io(ub)) { - __ublk_quiesce_dev(ub); - } else { - blk_mq_quiesce_queue(ub->ub_disk->queue); - ub->dev_info.state = UBLK_S_DEV_FAIL_IO; - for (i = 0; i < ub->dev_info.nr_hw_queues; i++) { - ublk_get_queue(ub, i)->fail_io = true; - } - blk_mq_unquiesce_queue(ub->ub_disk->queue); - } - - unlock: + ublk_stop_dev_unlocked(ub); mutex_unlock(&ub->mutex); ublk_cancel_dev(ub); } @@ -2419,7 +2447,6 @@ static int ublk_add_tag_set(struct ublk_device *ub) static void ublk_remove(struct ublk_device *ub) { ublk_stop_dev(ub); - cancel_work_sync(&ub->nosrv_work); cdev_device_del(&ub->cdev, &ub->cdev_dev); ublk_put_device(ub); ublks_added--; @@ -2693,7 +2720,6 @@ static int ublk_ctrl_add_dev(struct io_uring_cmd *cmd) goto out_unlock; mutex_init(&ub->mutex); spin_lock_init(&ub->lock); - INIT_WORK(&ub->nosrv_work, ublk_nosrv_work); ret = ublk_alloc_dev_number(ub, header->dev_id); if (ret < 0) @@ -2828,7 +2854,6 @@ static inline void ublk_ctrl_cmd_dump(struct io_uring_cmd *cmd) static int ublk_ctrl_stop_dev(struct ublk_device *ub) { ublk_stop_dev(ub); - cancel_work_sync(&ub->nosrv_work); return 0; } -- 2.43.0

2 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror