nvme-pci: move nvme_should_reset() to core code
PATCH v3 1/2 nvme-pci: move nvme_should_reset() to core code
Move nvme_should_reset() to core code so it fits with a subsequent core code patch. Tweak the interface to take a nvme_ctrl argument instead of nvme_dev, and fixup the one reference. No functional change.
---
Changes since v2: None
drivers/nvme/host/core.c | 19 +++++++++++++++++++ drivers/nvme/host/nvme.h | 1 + drivers/nvme/host/pci.c | 27 +-------------------------- 3 files changed, 21 insertions(+), 26 deletions(-)
drivers/nvme/host/core.c +19 -0
@@ -171,6 +171,25 @@ static inline void nvme_stop_failfast_work(struct nvme_ctrl *ctrl)clear_bit (NVME_CTRL_FAILFAST_EXPIRED , &ctrl->flags); } +bool nvme_should_reset (struct nvme_ctrl *ctrl,u32 csts) +{ +/* If there is a reset/reinit ongoing, we shouldn't reset again. */ +switch (ctrl->state) { +case NVME_CTRL_RESETTING : +case NVME_CTRL_CONNECTING : +return false; +default : +break ; + } + +/* +* We shouldn't reset unless the controller is on fatal error state +* or if we lost the communication with it. +*/ +return (csts &NVME_CSTS_CFS ) || + (ctrl->subsystem && (csts &NVME_CSTS_NSSRO )); +} +EXPORT_SYMBOL_GPL (nvme_should_reset);int nvme_reset_ctrl (struct nvme_ctrl *ctrl) {
drivers/nvme/host/nvme.h +1 -0
@@ -762,6 +762,7 @@ int nvme_get_features(struct nvme_ctrl *dev, unsigned int fid, u32 *result);int nvme_set_queue_count (struct nvme_ctrl *ctrl,int *count);void nvme_stop_keep_alive (struct nvme_ctrl *ctrl); +bool nvme_should_reset (struct nvme_ctrl *ctrl,u32 csts);int nvme_reset_ctrl (struct nvme_ctrl *ctrl);int nvme_reset_ctrl_sync (struct nvme_ctrl *ctrl);int nvme_try_sched_reset (struct nvme_ctrl *ctrl);
drivers/nvme/host/pci.c +1 -26
@@ -1293,31 +1293,6 @@ static void abort_endio(struct request *req, blk_status_t error)blk_mq_free_request (req); } -static bool nvme_should_reset (struct nvme_dev *dev,u32 csts) -{ -/* If true, indicates loss of adapter communication, possibly by a -* NVMe Subsystem reset. -*/ -bool nssro = dev->subsystem && (csts &NVME_CSTS_NSSRO ); - -/* If there is a reset/reinit ongoing, we shouldn't reset again. */ -switch (dev->ctrl.state) { -case NVME_CTRL_RESETTING : -case NVME_CTRL_CONNECTING : -return false; -default : -break ; - } - -/* We shouldn't reset unless the controller is on fatal error state -* _or_ if we lost the communication with it. -*/ -if (!(csts &NVME_CSTS_CFS ) && !nssro) -return false; - -return true; -} -static void nvme_warn_reset (struct nvme_dev *dev,u32 csts) {/* Read a config register to help see what died. */
@@ -1355,7 +1330,7 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req, bool reserved)/* * Reset immediately if the controller is failed */ -if (nvme_should_reset (dev, csts)) { +if (nvme_should_reset (&dev->ctrl, csts)) {nvme_warn_reset (dev, csts);nvme_dev_disable (dev, false);nvme_reset_ctrl (&dev->ctrl);
Signature · 1 line
1.8.3.1
PATCH v3 2/2 nvme: handle persistent internal error AER from NVMe controller
In the NVM Express Revision 1.4 spec, Figure 145 describes possible values for an AER with event type "Error" (value 000b). For a Persistent Internal Error (value 03h), the host should perform a controller reset.
Add support for this error using code that already exists for doing a controller reset. As part of this support, introduce two utility functions for parsing the AER type and subtype.
This new support was tested in a lab environment where we can generate the persistent internal error on demand, and observe both the Linux side and NVMe controller side to see that the controller reset has been done.
---
Changes since v2:
* Instead of reading CSTS, use a constant value as input to nvme_should_reset() [Keith Busch]
* Introduce helper functions for parsing the AER result fields [Keith Busch]
drivers/nvme/host/core.c | 41 +++++++++++++++++++++++++++++++++++++++-- include/linux/nvme.h | 4 ++++ 2 files changed, 43 insertions(+), 2 deletions(-)
drivers/nvme/host/core.c +39 -2
@@ -4521,9 +4521,19 @@ static void nvme_fw_act_work(struct work_struct *work)nvme_get_fw_slot_info (ctrl); } +static u32 nvme_aer_type (u32 result) +{ +return result &0x7 ; +} + +static u32 nvme_aer_subtype (u32 result) +{ +return (result &0xff00 ) >>8 ; +} +static void nvme_handle_aen_notice (struct nvme_ctrl *ctrl,u32 result) { -u32 aer_notice_type = (result &0xff00 ) >>8 ; +u32 aer_notice_type =nvme_aer_subtype (result);trace_nvme_async_event (ctrl, aer_notice_type);
@@ -4556,11 +4566,29 @@ static void nvme_handle_aen_notice(struct nvme_ctrl *ctrl, u32 result) } } +static void nvme_handle_aer_persistent_error (struct nvme_ctrl *ctrl) +{ +trace_nvme_async_event (ctrl,NVME_AER_ERROR ); + +/* +* We can't read the CSTS here because we're in an atomic context on +* some transports and the read may require submitting a request to the +* to the controller and getting a response. Such a sequence isn't +* likely to be successful anyway if the controller is reporting a +* persistent internal error. So assume CSTS.CFS is set. +*/ +if (nvme_should_reset (ctrl,NVME_CSTS_CFS )) { +dev_warn (ctrl->device,"resetting controller due to AER\n" ); +nvme_reset_ctrl (ctrl); + } +} +void nvme_complete_async_event (struct nvme_ctrl *ctrl,__le16 status,volatile union nvme_result *res) {u32 result =le32_to_cpu (res->u32); -u32 aer_type = result &0x07 ; +u32 aer_type =nvme_aer_type (result); +u32 aer_subtype =nvme_aer_subtype (result);if (le16_to_cpu (status) >>1 !=NVME_SC_SUCCESS )return ;
@@ -4570,6 +4598,15 @@ void nvme_complete_async_event(struct nvme_ctrl *ctrl, __le16 status,nvme_handle_aen_notice (ctrl, result);break ;case NVME_AER_ERROR : +/* +* For a persistent internal error, don't run async_event_work +* to submit a new AER. The controller reset will do it. +*/ +if (aer_subtype ==NVME_AER_ERROR_PERSIST_INT_ERR ) { +nvme_handle_aer_persistent_error (ctrl); +return ; + } + fallthrough;case NVME_AER_SMART :case NVME_AER_CSS :case NVME_AER_VS :
include/linux/nvme.h +4 -0
@@ -712,6 +712,10 @@ enum { };enum { +NVME_AER_ERROR_PERSIST_INT_ERR =0x03 , +}; + +enum {NVME_AER_NOTICE_NS_CHANGED =0x00 ,NVME_AER_NOTICE_FW_ACT_STARTING =0x01 ,NVME_AER_NOTICE_ANA =0x03 ,
Signature · 1 line
1.8.3.1
Christoph Hellwig
maintainerRe: [PATCH v3 2/2] nvme: handle persistent internal error AER from NVMe controller
Quoted 4 lines · Michael Kelley
drivers/nvme/host/core.c +4
-0+static void nvme_handle_aer_persistent_error (struct nvme_ctrl *ctrl) +{ +trace_nvme_async_event (ctrl,NVME_AER_ERROR ); +
drivers/nvme/host/core.c static void nvme_handle_aen_notice(struct nvme_ctrl *ctrl, u32 result) · line 4582
Quoted 10 lines
drivers/nvme/host/core.c +10
-0+/* +* We can't read the CSTS here because we're in an atomic context on +* some transports and the read may require submitting a request to the +* to the controller and getting a response. Such a sequence isn't +* likely to be successful anyway if the controller is reporting a +* persistent internal error. So assume CSTS.CFS is set. +*/ +if (nvme_should_reset (ctrl,NVME_CSTS_CFS )) { +dev_warn (ctrl->device,"resetting controller due to AER\n" ); +nvme_reset_ctrl (ctrl);
I don't think we even need the nvme_should_reset check now.
nvme_reset_ctrl first calls nvme_change_ctrl_state, which only allows the transition to the RESETTING state if it previously was NEW or LIVE, so we are already covered. The only downside would be an extra kernel message if we already were in another state.
RE: [PATCH v3 2/2] nvme: handle persistent internal error AER from NVMe controller
From: Christoph Hellwig <[email protected]> Sent: Tuesday, June 7, 2022 3:36 AM
Quoted 19 lines · Michael Kelley
Quoted 4 lines
drivers/nvme/host/core.c +4
-0+static void nvme_handle_aer_persistent_error (struct nvme_ctrl *ctrl) +{ +trace_nvme_async_event (ctrl,NVME_AER_ERROR ); +drivers/nvme/host/core.c static void nvme_handle_aen_notice(struct nvme_ctrl *ctrl, u32 result) · line 4582
Quoted 10 lines
drivers/nvme/host/core.c +10
-0+/* +* We can't read the CSTS here because we're in an atomic context on +* some transports and the read may require submitting a request to the +* to the controller and getting a response. Such a sequence isn't +* likely to be successful anyway if the controller is reporting a +* persistent internal error. So assume CSTS.CFS is set. +*/ +if (nvme_should_reset (ctrl,NVME_CSTS_CFS )) { +dev_warn (ctrl->device,"resetting controller due to AER\n" ); +nvme_reset_ctrl (ctrl);I don't think we even need the
nvme_should_resetcheck now.
Quoted 4 lines
nvme_reset_ctrlfirst callsnvme_change_ctrl_state, which only allows the transition to the RESETTING state if it previously was NEW or LIVE, so we are already covered. The only downside would be an extra kernel message if we already were in another state.
OK, I agree. Patch 1/2 can be dropped since there's now no need to move nvme_should_reset(), and patch 2 is simplified even further.
I'll do a v4.
Michael