From 1a878284473284f9577d44babf16d87152a05c33 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 29 Jul 2011 17:16:40 -0700 Subject: [PATCH 01/25] [SCSI] isci: fix sata response handling A bug (likely copy/paste) that has been carried from the original implementation. The unsolicited frame handling structure returns the d2h fis in the isci_request.stp.rsp buffer. Cc: Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/request.c | 18 ++++++------------ 1 file changed, 6 insertions(+), 12 deletions(-) diff --git a/drivers/scsi/isci/request.c b/drivers/scsi/isci/request.c index a46e07ac789f..b4cf998385b3 100644 --- a/drivers/scsi/isci/request.c +++ b/drivers/scsi/isci/request.c @@ -2399,22 +2399,19 @@ static void isci_task_save_for_upper_layer_completion( } } -static void isci_request_process_stp_response(struct sas_task *task, - void *response_buffer) +static void isci_process_stp_response(struct sas_task *task, struct dev_to_host_fis *fis) { - struct dev_to_host_fis *d2h_reg_fis = response_buffer; struct task_status_struct *ts = &task->task_status; struct ata_task_resp *resp = (void *)&ts->buf[0]; - resp->frame_len = le16_to_cpu(*(__le16 *)(response_buffer + 6)); - memcpy(&resp->ending_fis[0], response_buffer + 16, 24); + resp->frame_len = sizeof(*fis); + memcpy(resp->ending_fis, fis, sizeof(*fis)); ts->buf_valid_size = sizeof(*resp); - /** - * If the device fault bit is set in the status register, then + /* If the device fault bit is set in the status register, then * set the sense data and return. */ - if (d2h_reg_fis->status & ATA_DF) + if (fis->status & ATA_DF) ts->stat = SAS_PROTO_RESPONSE; else ts->stat = SAM_STAT_GOOD; @@ -2428,7 +2425,6 @@ static void isci_request_io_request_complete(struct isci_host *ihost, { struct sas_task *task = isci_request_access_task(request); struct ssp_response_iu *resp_iu; - void *resp_buf; unsigned long task_flags; struct isci_remote_device *idev = isci_lookup_device(task->dev); enum service_response response = SAS_TASK_UNDELIVERED; @@ -2565,9 +2561,7 @@ static void isci_request_io_request_complete(struct isci_host *ihost, task); if (sas_protocol_ata(task->task_proto)) { - resp_buf = &request->stp.rsp; - isci_request_process_stp_response(task, - resp_buf); + isci_process_stp_response(task, &request->stp.rsp); } else if (SAS_PROTOCOL_SSP == task->task_proto) { /* crack the iu response buffer. */ From ee33e2b771f9e9e4aaba2bb2ace7b727fe451a8b Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 29 Jul 2011 17:16:45 -0700 Subject: [PATCH 02/25] [SCSI] isci: fix 32-bit operation when CONFIG_HIGHMEM64G=n The unsolicited frame control infrastructure requires a table of dma addresses for the hardware to lookup the frame buffer location by an index. The hardware expects the elements of this table to be 64-bit quantities, so we cannot reference these elements as dma_addr_t. All unsolicited frame protocols are affected, particularly SATA-PIO and SMP which prevented direct-attached SATA drives and expander-attached drives to not be discovered. Cc: Reported-by: Jacek Danecki Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/unsolicited_frame_control.c | 2 +- drivers/scsi/isci/unsolicited_frame_control.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/isci/unsolicited_frame_control.c b/drivers/scsi/isci/unsolicited_frame_control.c index e9e1e2abacb9..16f88ab939c8 100644 --- a/drivers/scsi/isci/unsolicited_frame_control.c +++ b/drivers/scsi/isci/unsolicited_frame_control.c @@ -72,7 +72,7 @@ int sci_unsolicited_frame_control_construct(struct isci_host *ihost) */ buf_len = SCU_MAX_UNSOLICITED_FRAMES * SCU_UNSOLICITED_FRAME_BUFFER_SIZE; header_len = SCU_MAX_UNSOLICITED_FRAMES * sizeof(struct scu_unsolicited_frame_header); - size = buf_len + header_len + SCU_MAX_UNSOLICITED_FRAMES * sizeof(dma_addr_t); + size = buf_len + header_len + SCU_MAX_UNSOLICITED_FRAMES * sizeof(uf_control->address_table.array[0]); /* * The Unsolicited Frame buffers are set at the start of the UF diff --git a/drivers/scsi/isci/unsolicited_frame_control.h b/drivers/scsi/isci/unsolicited_frame_control.h index 31cb9506f52d..75d896686f5a 100644 --- a/drivers/scsi/isci/unsolicited_frame_control.h +++ b/drivers/scsi/isci/unsolicited_frame_control.h @@ -214,7 +214,7 @@ struct sci_uf_address_table_array { * starting address of the UF address table. * 64-bit pointers are required by the hardware. */ - dma_addr_t *array; + u64 *array; /** * This field specifies the physical address location for the UF From 985af6f70dbb8a33b3af8a7c7df508d924650e37 Mon Sep 17 00:00:00 2001 From: Marcin Tomczak Date: Fri, 29 Jul 2011 17:16:50 -0700 Subject: [PATCH 03/25] [SCSI] isci: change sas phy timeouts from 54us to 59us Need the following workaround in the driver for interoperability with the older Intel SSD drives and any other SATA drive that may exhibit the same behavior. This is a corner case where SCU speed is limited to either 3G or 1.5G and the drive has a period of DC idle when it switches speed during SATA speed negotiation. Workaround :change PHYTOV[31:24] from 0x36 to 0x3B. Signed-off-by: Marcin Tomczak Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/phy.c | 13 +++++++++++++ drivers/scsi/isci/registers.h | 12 ++++++++++++ 2 files changed, 25 insertions(+) diff --git a/drivers/scsi/isci/phy.c b/drivers/scsi/isci/phy.c index 79313a7a2356..430fc8ff014a 100644 --- a/drivers/scsi/isci/phy.c +++ b/drivers/scsi/isci/phy.c @@ -104,6 +104,7 @@ sci_phy_link_layer_initialization(struct isci_phy *iphy, u32 parity_count = 0; u32 llctl, link_rate; u32 clksm_value = 0; + u32 sp_timeouts = 0; iphy->link_layer_registers = reg; @@ -211,6 +212,18 @@ sci_phy_link_layer_initialization(struct isci_phy *iphy, llctl |= SCU_SAS_LLCTL_GEN_VAL(MAX_LINK_RATE, link_rate); writel(llctl, &iphy->link_layer_registers->link_layer_control); + sp_timeouts = readl(&iphy->link_layer_registers->sas_phy_timeouts); + + /* Clear the default 0x36 (54us) RATE_CHANGE timeout value. */ + sp_timeouts &= ~SCU_SAS_PHYTOV_GEN_VAL(RATE_CHANGE, 0xFF); + + /* Set RATE_CHANGE timeout value to 0x3B (59us). This ensures SCU can + * lock with 3Gb drive when SCU max rate is set to 1.5Gb. + */ + sp_timeouts |= SCU_SAS_PHYTOV_GEN_VAL(RATE_CHANGE, 0x3B); + + writel(sp_timeouts, &iphy->link_layer_registers->sas_phy_timeouts); + if (is_a2(ihost->pdev)) { /* Program the max ARB time for the PHY to 700us so we inter-operate with * the PMC expander which shuts down PHYs if the expander PHY generates too diff --git a/drivers/scsi/isci/registers.h b/drivers/scsi/isci/registers.h index 9b266c7428e8..00afc738bbed 100644 --- a/drivers/scsi/isci/registers.h +++ b/drivers/scsi/isci/registers.h @@ -1299,6 +1299,18 @@ struct scu_transport_layer_registers { #define SCU_AFE_XCVRCR_OFFSET 0x00DC #define SCU_AFE_LUTCR_OFFSET 0x00E0 +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_ALIGN_DETECTION_SHIFT (0UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_ALIGN_DETECTION_MASK (0x000000FFUL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_HOT_PLUG_SHIFT (8UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_HOT_PLUG_MASK (0x0000FF00UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_COMSAS_DETECTION_SHIFT (16UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_COMSAS_DETECTION_MASK (0x00FF0000UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_RATE_CHANGE_SHIFT (24UL) +#define SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_RATE_CHANGE_MASK (0xFF000000UL) + +#define SCU_SAS_PHYTOV_GEN_VAL(name, value) \ + SCU_GEN_VALUE(SCU_SAS_PHY_TIMER_TIMEOUT_VALUES_##name, value) + #define SCU_SAS_LINK_LAYER_CONTROL_MAX_LINK_RATE_SHIFT (0) #define SCU_SAS_LINK_LAYER_CONTROL_MAX_LINK_RATE_MASK (0x00000003) #define SCU_SAS_LINK_LAYER_CONTROL_MAX_LINK_RATE_GEN1 (0) From 4ac13e177904280a2502c27029a72e3fd2957cde Mon Sep 17 00:00:00 2001 From: Dave Jiang Date: Fri, 29 Jul 2011 17:16:55 -0700 Subject: [PATCH 04/25] [SCSI] isci: Update MAINTAINERS entry for the isci driver Signed-off-by: Dave Jiang Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- MAINTAINERS | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 069ee3b5c651..4fe6854c57d9 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -3262,6 +3262,17 @@ F: Documentation/input/multi-touch-protocol.txt F: drivers/input/input-mt.c K: \b(ABS|SYN)_MT_ +INTEL C600 SERIES SAS CONTROLLER DRIVER +M: Intel SCU Linux support +M: Dan Williams +M: Dave Jiang +M: Ed Nadolski +L: linux-scsi@vger.kernel.org +T: git git://git.kernel.org/pub/scm/linux/kernel/git/djbw/isci.git +S: Maintained +F: drivers/scsi/isci/ +F: firmware/isci/ + INTEL IDLE DRIVER M: Len Brown L: linux-pm@lists.linux-foundation.org From 3a7bda830fad427768ed71c0ebf3448849c006b5 Mon Sep 17 00:00:00 2001 From: Dave Jiang Date: Fri, 29 Jul 2011 17:17:00 -0700 Subject: [PATCH 05/25] [SCSI] isci: Adding documentation to API change and fixup sysfs registration Adding API update for adding isci_id entry scsi_host sysfs entry. Also fixing up the sysfs registration to the scsi_host template Signed-off-by: Dave Jiang Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- .../ABI/testing/sysfs-class-scsi_host | 13 +++++++ drivers/scsi/isci/init.c | 36 +++++++++---------- 2 files changed, 31 insertions(+), 18 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-class-scsi_host diff --git a/Documentation/ABI/testing/sysfs-class-scsi_host b/Documentation/ABI/testing/sysfs-class-scsi_host new file mode 100644 index 000000000000..29a4f892e433 --- /dev/null +++ b/Documentation/ABI/testing/sysfs-class-scsi_host @@ -0,0 +1,13 @@ +What: /sys/class/scsi_host/hostX/isci_id +Date: June 2011 +Contact: Dave Jiang +Description: + This file contains the enumerated host ID for the Intel + SCU controller. The Intel(R) C600 Series Chipset SATA/SAS + Storage Control Unit embeds up to two 4-port controllers in + a single PCI device. The controllers are enumerated in order + which usually means the lowest number scsi_host corresponds + with the first controller, but this association is not + guaranteed. The 'isci_id' attribute unambiguously identifies + the controller index: '0' for the first controller, + '1' for the second. diff --git a/drivers/scsi/isci/init.c b/drivers/scsi/isci/init.c index 61e0d09e2b57..e78320bbec4f 100644 --- a/drivers/scsi/isci/init.c +++ b/drivers/scsi/isci/init.c @@ -59,6 +59,7 @@ #include #include #include +#include #include "isci.h" #include "task.h" #include "probe_roms.h" @@ -113,6 +114,22 @@ unsigned char max_concurr_spinup = 1; module_param(max_concurr_spinup, byte, 0); MODULE_PARM_DESC(max_concurr_spinup, "Max concurrent device spinup"); +static ssize_t isci_show_id(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct Scsi_Host *shost = container_of(dev, typeof(*shost), shost_dev); + struct sas_ha_struct *sas_ha = SHOST_TO_SAS_HA(shost); + struct isci_host *ihost = container_of(sas_ha, typeof(*ihost), sas_ha); + + return snprintf(buf, PAGE_SIZE, "%d\n", ihost->id); +} + +static DEVICE_ATTR(isci_id, S_IRUGO, isci_show_id, NULL); + +struct device_attribute *isci_host_attrs[] = { + &dev_attr_isci_id, + NULL +}; + static struct scsi_host_template isci_sht = { .module = THIS_MODULE, @@ -138,6 +155,7 @@ static struct scsi_host_template isci_sht = { .slave_alloc = sas_slave_alloc, .target_destroy = sas_target_destroy, .ioctl = sas_ioctl, + .shost_attrs = isci_host_attrs, }; static struct sas_domain_function_template isci_transport_ops = { @@ -232,17 +250,6 @@ static int isci_register_sas_ha(struct isci_host *isci_host) return 0; } -static ssize_t isci_show_id(struct device *dev, struct device_attribute *attr, char *buf) -{ - struct Scsi_Host *shost = container_of(dev, typeof(*shost), shost_dev); - struct sas_ha_struct *sas_ha = SHOST_TO_SAS_HA(shost); - struct isci_host *ihost = container_of(sas_ha, typeof(*ihost), sas_ha); - - return snprintf(buf, PAGE_SIZE, "%d\n", ihost->id); -} - -static DEVICE_ATTR(isci_id, S_IRUGO, isci_show_id, NULL); - static void isci_unregister(struct isci_host *isci_host) { struct Scsi_Host *shost; @@ -251,7 +258,6 @@ static void isci_unregister(struct isci_host *isci_host) return; shost = isci_host->shost; - device_remove_file(&shost->shost_dev, &dev_attr_isci_id); sas_unregister_ha(&isci_host->sas_ha); @@ -415,14 +421,8 @@ static struct isci_host *isci_host_alloc(struct pci_dev *pdev, int id) if (err) goto err_shost_remove; - err = device_create_file(&shost->shost_dev, &dev_attr_isci_id); - if (err) - goto err_unregister_ha; - return isci_host; - err_unregister_ha: - sas_unregister_ha(&(isci_host->sas_ha)); err_shost_remove: scsi_remove_host(shost); err_shost: From 39ea2c5b5ffaa344467da53e885cfa4ac0105050 Mon Sep 17 00:00:00 2001 From: Jeff Skirvin Date: Fri, 29 Jul 2011 17:17:05 -0700 Subject: [PATCH 06/25] [SCSI] isci: Leave requests alone if already terminating. Instead of immediately completing any request that has a second termination call made on it, wait for the TC done/abort HW event. Signed-off-by: Jeff Skirvin Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/request.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/isci/request.c b/drivers/scsi/isci/request.c index b4cf998385b3..b5d3a8c4d329 100644 --- a/drivers/scsi/isci/request.c +++ b/drivers/scsi/isci/request.c @@ -732,12 +732,20 @@ sci_io_request_terminate(struct isci_request *ireq) sci_change_state(&ireq->sm, SCI_REQ_ABORTING); return SCI_SUCCESS; case SCI_REQ_TASK_WAIT_TC_RESP: + /* The task frame was already confirmed to have been + * sent by the SCU HW. Since the state machine is + * now only waiting for the task response itself, + * abort the request and complete it immediately + * and don't wait for the task response. + */ sci_change_state(&ireq->sm, SCI_REQ_ABORTING); sci_change_state(&ireq->sm, SCI_REQ_COMPLETED); return SCI_SUCCESS; case SCI_REQ_ABORTING: - sci_change_state(&ireq->sm, SCI_REQ_COMPLETED); - return SCI_SUCCESS; + /* If a request has a termination requested twice, return + * a failure indication, since HW confirmation of the first + * abort is still outstanding. + */ case SCI_REQ_COMPLETED: default: dev_warn(&ireq->owning_controller->pdev->dev, From 9b4be528999483d70a1ffc0accd102e477d5a503 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 29 Jul 2011 17:17:10 -0700 Subject: [PATCH 07/25] [SCSI] isci: dynamic interrupt coalescing Hardware allows both an outstanding number commands and a timeout value (whichever occurs first) as a gate to the next interrupt generation. This scheme at completion time looks at the remaining number of outstanding tasks and sets the timeout to maximize small transaction operation. If transactions are large (take more than a few 10s of microseconds to complete) then performance is not interrupt processing bound, so the small timeouts this scheme generates are overridden by the time it takes for a completion to arrive. Tested-by: Dave Jiang Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/host.c | 10 +++++++++- drivers/scsi/isci/host.h | 3 +++ 2 files changed, 12 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/isci/host.c b/drivers/scsi/isci/host.c index 26072f1e9852..2328f98c7f1e 100644 --- a/drivers/scsi/isci/host.c +++ b/drivers/scsi/isci/host.c @@ -1091,6 +1091,7 @@ static void isci_host_completion_routine(unsigned long data) struct isci_request *request; struct isci_request *next_request; struct sas_task *task; + u16 active; INIT_LIST_HEAD(&completed_request_list); INIT_LIST_HEAD(&errored_request_list); @@ -1181,6 +1182,13 @@ static void isci_host_completion_routine(unsigned long data) } } + /* the coalesence timeout doubles at each encoding step, so + * update it based on the ilog2 value of the outstanding requests + */ + active = isci_tci_active(ihost); + writel(SMU_ICC_GEN_VAL(NUMBER, active) | + SMU_ICC_GEN_VAL(TIMER, ISCI_COALESCE_BASE + ilog2(active)), + &ihost->smu_registers->interrupt_coalesce_control); } /** @@ -1471,7 +1479,7 @@ static void sci_controller_ready_state_enter(struct sci_base_state_machine *sm) struct isci_host *ihost = container_of(sm, typeof(*ihost), sm); /* set the default interrupt coalescence number and timeout value. */ - sci_controller_set_interrupt_coalescence(ihost, 0x10, 250); + sci_controller_set_interrupt_coalescence(ihost, 0, 0); } static void sci_controller_ready_state_exit(struct sci_base_state_machine *sm) diff --git a/drivers/scsi/isci/host.h b/drivers/scsi/isci/host.h index 062101a39f79..9f33831a2f04 100644 --- a/drivers/scsi/isci/host.h +++ b/drivers/scsi/isci/host.h @@ -369,6 +369,9 @@ static inline struct isci_host *dev_to_ihost(struct domain_device *dev) #define ISCI_TAG_SEQ(tag) (((tag) >> 12) & (SCI_MAX_SEQ-1)) #define ISCI_TAG_TCI(tag) ((tag) & (SCI_MAX_IO_REQUESTS-1)) +/* interrupt coalescing baseline: 9 == 3 to 5us interrupt delay per command */ +#define ISCI_COALESCE_BASE 9 + /* expander attached sata devices require 3 rnc slots */ static inline int sci_remote_device_node_count(struct isci_remote_device *idev) { From 77cd72a53f6426f81b7f56a862402849ee903bda Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 29 Jul 2011 17:17:16 -0700 Subject: [PATCH 08/25] [SCSI] isci: fix event-get pointer increment Hardware only increments the put pointer on event types >= 4. Do not increment the get pointer for event type 3. Reported-by: Kapil Karkra Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/host.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/scsi/isci/host.c b/drivers/scsi/isci/host.c index 2328f98c7f1e..6981b773a88d 100644 --- a/drivers/scsi/isci/host.c +++ b/drivers/scsi/isci/host.c @@ -531,6 +531,9 @@ static void sci_controller_process_completions(struct isci_host *ihost) break; case SCU_COMPLETION_TYPE_EVENT: + sci_controller_event_completion(ihost, ent); + break; + case SCU_COMPLETION_TYPE_NOTIFY: { event_cycle ^= ((event_get+1) & SCU_MAX_EVENTS) << (SMU_COMPLETION_QUEUE_GET_EVENT_CYCLE_BIT_SHIFT - SCU_MAX_EVENTS_SHIFT); From 98e2a5a3a125608505783bdb95744997f76b3c30 Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Fri, 29 Jul 2011 17:17:21 -0700 Subject: [PATCH 09/25] [SCSI] isci: add version number Signed-off-by: Dan Williams Signed-off-by: James Bottomley --- drivers/scsi/isci/init.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/isci/init.c b/drivers/scsi/isci/init.c index e78320bbec4f..29aa34efb0f5 100644 --- a/drivers/scsi/isci/init.c +++ b/drivers/scsi/isci/init.c @@ -64,6 +64,14 @@ #include "task.h" #include "probe_roms.h" +#define MAJ 1 +#define MIN 0 +#define BUILD 0 +#define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \ + __stringify(BUILD) + +MODULE_VERSION(DRV_VERSION); + static struct scsi_transport_template *isci_transport_template; static DEFINE_PCI_DEVICE_TABLE(isci_id_table) = { @@ -540,7 +548,8 @@ static __init int isci_init(void) { int err; - pr_info("%s: Intel(R) C600 SAS Controller Driver\n", DRV_NAME); + pr_info("%s: Intel(R) C600 SAS Controller Driver - version %s\n", + DRV_NAME, DRV_VERSION); isci_transport_template = sas_domain_attach_transport(&isci_transport_ops); if (!isci_transport_template) From 0b0e1d6cbcc8627970e0399df8f06edd690ec7d9 Mon Sep 17 00:00:00 2001 From: "Stephen M. Cameron" Date: Tue, 9 Aug 2011 08:17:30 -0500 Subject: [PATCH 10/25] [SCSI] hpsa: fix problem that OBDR devices are not detected The test to detect OBDR ("One Button Disaster Recovery") cd-rom devices was comparing against uninitialized data. Fixed by moving the test for the device to where the inquiry data is collected, and uninitialized variable altogether as it wasn't really being used. Signed-off-by: Stephen M. Cameron Cc: stable@kernel.org Signed-off-by: James Bottomley --- drivers/scsi/hpsa.c | 47 ++++++++++++++++++++++++++------------------- 1 file changed, 27 insertions(+), 20 deletions(-) diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c index ec61bdb833ac..1f32f0610bc0 100644 --- a/drivers/scsi/hpsa.c +++ b/drivers/scsi/hpsa.c @@ -1548,10 +1548,17 @@ static inline void hpsa_set_bus_target_lun(struct hpsa_scsi_dev_t *device, } static int hpsa_update_device_info(struct ctlr_info *h, - unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device) + unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device, + unsigned char *is_OBDR_device) { -#define OBDR_TAPE_INQ_SIZE 49 + +#define OBDR_SIG_OFFSET 43 +#define OBDR_TAPE_SIG "$DR-10" +#define OBDR_SIG_LEN (sizeof(OBDR_TAPE_SIG) - 1) +#define OBDR_TAPE_INQ_SIZE (OBDR_SIG_OFFSET + OBDR_SIG_LEN) + unsigned char *inq_buff; + unsigned char *obdr_sig; inq_buff = kzalloc(OBDR_TAPE_INQ_SIZE, GFP_KERNEL); if (!inq_buff) @@ -1583,6 +1590,16 @@ static int hpsa_update_device_info(struct ctlr_info *h, else this_device->raid_level = RAID_UNKNOWN; + if (is_OBDR_device) { + /* See if this is a One-Button-Disaster-Recovery device + * by looking for "$DR-10" at offset 43 in inquiry data. + */ + obdr_sig = &inq_buff[OBDR_SIG_OFFSET]; + *is_OBDR_device = (this_device->devtype == TYPE_ROM && + strncmp(obdr_sig, OBDR_TAPE_SIG, + OBDR_SIG_LEN) == 0); + } + kfree(inq_buff); return 0; @@ -1716,7 +1733,7 @@ static int add_msa2xxx_enclosure_device(struct ctlr_info *h, return 0; } - if (hpsa_update_device_info(h, scsi3addr, this_device)) + if (hpsa_update_device_info(h, scsi3addr, this_device, NULL)) return 0; (*nmsa2xxx_enclosures)++; hpsa_set_bus_target_lun(this_device, bus, target, 0); @@ -1808,7 +1825,6 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) */ struct ReportLUNdata *physdev_list = NULL; struct ReportLUNdata *logdev_list = NULL; - unsigned char *inq_buff = NULL; u32 nphysicals = 0; u32 nlogicals = 0; u32 ndev_allocated = 0; @@ -1824,11 +1840,9 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) GFP_KERNEL); physdev_list = kzalloc(reportlunsize, GFP_KERNEL); logdev_list = kzalloc(reportlunsize, GFP_KERNEL); - inq_buff = kmalloc(OBDR_TAPE_INQ_SIZE, GFP_KERNEL); tmpdevice = kzalloc(sizeof(*tmpdevice), GFP_KERNEL); - if (!currentsd || !physdev_list || !logdev_list || - !inq_buff || !tmpdevice) { + if (!currentsd || !physdev_list || !logdev_list || !tmpdevice) { dev_err(&h->pdev->dev, "out of memory\n"); goto out; } @@ -1863,7 +1877,7 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) /* adjust our table of devices */ nmsa2xxx_enclosures = 0; for (i = 0; i < nphysicals + nlogicals + 1; i++) { - u8 *lunaddrbytes; + u8 *lunaddrbytes, is_OBDR = 0; /* Figure out where the LUN ID info is coming from */ lunaddrbytes = figure_lunaddrbytes(h, raid_ctlr_position, @@ -1874,7 +1888,8 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) continue; /* Get device type, vendor, model, device id */ - if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice)) + if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice, + &is_OBDR)) continue; /* skip it if we can't talk to it. */ figure_bus_target_lun(h, lunaddrbytes, &bus, &target, &lun, tmpdevice); @@ -1898,7 +1913,7 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) hpsa_set_bus_target_lun(this_device, bus, target, lun); switch (this_device->devtype) { - case TYPE_ROM: { + case TYPE_ROM: /* We don't *really* support actual CD-ROM devices, * just "One Button Disaster Recovery" tape drive * which temporarily pretends to be a CD-ROM drive. @@ -1906,15 +1921,8 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) * device by checking for "$DR-10" in bytes 43-48 of * the inquiry data. */ - char obdr_sig[7]; -#define OBDR_TAPE_SIG "$DR-10" - strncpy(obdr_sig, &inq_buff[43], 6); - obdr_sig[6] = '\0'; - if (strncmp(obdr_sig, OBDR_TAPE_SIG, 6) != 0) - /* Not OBDR device, ignore it. */ - break; - } - ncurrent++; + if (is_OBDR) + ncurrent++; break; case TYPE_DISK: if (i < nphysicals) @@ -1947,7 +1955,6 @@ out: for (i = 0; i < ndev_allocated; i++) kfree(currentsd[i]); kfree(currentsd); - kfree(inq_buff); kfree(physdev_list); kfree(logdev_list); } From 01350d05539d1c95ef3568d062d864ab76ae7670 Mon Sep 17 00:00:00 2001 From: "Stephen M. Cameron" Date: Tue, 9 Aug 2011 08:18:01 -0500 Subject: [PATCH 11/25] [SCSI] hpsa: fix physical device lun and target numbering problem If a physical device exposed to the OS by hpsa is replaced (e.g. one hot plug tape drive is replaced by another, or a tape drive is placed into "OBDR" mode in which it acts like a CD-ROM device) and a rescan is initiated, the replaced device will be added to the SCSI midlayer with target and lun numbers set to -1. After that, a panic is likely to ensue. When a physical device is replaced, the lun and target number should be preserved. Signed-off-by: Stephen M. Cameron Cc: stable@kernel.org Signed-off-by: James Bottomley --- drivers/scsi/hpsa.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c index 1f32f0610bc0..b200b736b000 100644 --- a/drivers/scsi/hpsa.c +++ b/drivers/scsi/hpsa.c @@ -676,6 +676,16 @@ static void hpsa_scsi_replace_entry(struct ctlr_info *h, int hostno, BUG_ON(entry < 0 || entry >= HPSA_MAX_SCSI_DEVS_PER_HBA); removed[*nremoved] = h->dev[entry]; (*nremoved)++; + + /* + * New physical devices won't have target/lun assigned yet + * so we need to preserve the values in the slot we are replacing. + */ + if (new_entry->target == -1) { + new_entry->target = h->dev[entry]->target; + new_entry->lun = h->dev[entry]->lun; + } + h->dev[entry] = new_entry; added[*nadded] = new_entry; (*nadded)++; From 8cb2049c744809193ed3707a37c09676a24599ee Mon Sep 17 00:00:00 2001 From: Arun Easi Date: Tue, 16 Aug 2011 11:29:22 -0700 Subject: [PATCH 12/25] [SCSI] qla2xxx: T10 DIF - Handle uninitalized sectors. Driver needs to update protection bytes for uninitialized sectors as they are not DMA-d. Signed-off-by: Arun Easi Reviewed-by: Andrew Vasquez Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_attr.c | 5 +- drivers/scsi/qla2xxx/qla_fw.h | 5 + drivers/scsi/qla2xxx/qla_inline.h | 21 +++ drivers/scsi/qla2xxx/qla_iocb.c | 233 ++++++++++++++++++++++++++++-- drivers/scsi/qla2xxx/qla_isr.c | 90 ++++++++++-- drivers/scsi/qla2xxx/qla_os.c | 19 ++- 6 files changed, 335 insertions(+), 38 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c index 7836eb01c7fc..810067099801 100644 --- a/drivers/scsi/qla2xxx/qla_attr.c +++ b/drivers/scsi/qla2xxx/qla_attr.c @@ -1788,11 +1788,14 @@ qla24xx_vport_create(struct fc_vport *fc_vport, bool disable) if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) { if (ha->fw_attributes & BIT_4) { + int prot = 0; vha->flags.difdix_supported = 1; ql_dbg(ql_dbg_user, vha, 0x7082, "Registered for DIF/DIX type 1 and 3 protection.\n"); + if (ql2xenabledif == 1) + prot = SHOST_DIX_TYPE0_PROTECTION; scsi_host_set_prot(vha->host, - SHOST_DIF_TYPE1_PROTECTION + prot | SHOST_DIF_TYPE1_PROTECTION | SHOST_DIF_TYPE2_PROTECTION | SHOST_DIF_TYPE3_PROTECTION | SHOST_DIX_TYPE1_PROTECTION diff --git a/drivers/scsi/qla2xxx/qla_fw.h b/drivers/scsi/qla2xxx/qla_fw.h index 691783abfb69..aa69486dc064 100644 --- a/drivers/scsi/qla2xxx/qla_fw.h +++ b/drivers/scsi/qla2xxx/qla_fw.h @@ -537,6 +537,11 @@ struct sts_entry_24xx { /* * If DIF Error is set in comp_status, these additional fields are * defined: + * + * !!! NOTE: Firmware sends expected/actual DIF data in big endian + * format; but all of the "data" field gets swab32-d in the beginning + * of qla2x00_status_entry(). + * * &data[10] : uint8_t report_runt_bg[2]; - computed guard * &data[12] : uint8_t actual_dif[8]; - DIF Data received * &data[20] : uint8_t expected_dif[8]; - DIF Data computed diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h index d2e904bc21c0..c06e5f9b431e 100644 --- a/drivers/scsi/qla2xxx/qla_inline.h +++ b/drivers/scsi/qla2xxx/qla_inline.h @@ -102,3 +102,24 @@ qla2x00_set_fcport_state(fc_port_t *fcport, int state) fcport->d_id.b.al_pa); } } + +static inline int +qla2x00_hba_err_chk_enabled(unsigned char op) +{ + switch (op) { + case SCSI_PROT_READ_STRIP: + case SCSI_PROT_WRITE_INSERT: + if (ql2xenablehba_err_chk >= 1) + return 1; + break; + case SCSI_PROT_READ_PASS: + case SCSI_PROT_WRITE_PASS: + if (ql2xenablehba_err_chk >= 2) + return 1; + break; + case SCSI_PROT_READ_INSERT: + case SCSI_PROT_WRITE_STRIP: + return 1; + } + return 0; +} diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c index 49d6906af886..09ad3ce60064 100644 --- a/drivers/scsi/qla2xxx/qla_iocb.c +++ b/drivers/scsi/qla2xxx/qla_iocb.c @@ -717,12 +717,17 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, unsigned char op = scsi_get_prot_op(cmd); switch (scsi_get_prot_type(cmd)) { - /* For TYPE 0 protection: no checking */ case SCSI_PROT_DIF_TYPE0: - pkt->ref_tag_mask[0] = 0x00; - pkt->ref_tag_mask[1] = 0x00; - pkt->ref_tag_mask[2] = 0x00; - pkt->ref_tag_mask[3] = 0x00; + /* + * No check for ql2xenablehba_err_chk, as it would be an + * I/O error if hba tag generation is not done. + */ + pkt->ref_tag = cpu_to_le32((uint32_t) + (0xffffffff & scsi_get_lba(cmd))); + pkt->ref_tag_mask[0] = 0xff; + pkt->ref_tag_mask[1] = 0xff; + pkt->ref_tag_mask[2] = 0xff; + pkt->ref_tag_mask[3] = 0xff; break; /* @@ -730,7 +735,7 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, * match LBA in CDB + N */ case SCSI_PROT_DIF_TYPE2: - if (!ql2xenablehba_err_chk) + if (!qla2x00_hba_err_chk_enabled(op)) break; if (scsi_prot_sg_count(cmd)) { @@ -763,7 +768,7 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, * 16 bit app tag. */ case SCSI_PROT_DIF_TYPE1: - if (!ql2xenablehba_err_chk) + if (!qla2x00_hba_err_chk_enabled(op)) break; if (protcnt && (op == SCSI_PROT_WRITE_STRIP || @@ -798,7 +803,161 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, scsi_get_prot_type(cmd), cmd); } +struct qla2_sgx { + dma_addr_t dma_addr; /* OUT */ + uint32_t dma_len; /* OUT */ + uint32_t tot_bytes; /* IN */ + struct scatterlist *cur_sg; /* IN */ + + /* for book keeping, bzero on initial invocation */ + uint32_t bytes_consumed; + uint32_t num_bytes; + uint32_t tot_partial; + + /* for debugging */ + uint32_t num_sg; + srb_t *sp; +}; + +static int +qla24xx_get_one_block_sg(uint32_t blk_sz, struct qla2_sgx *sgx, + uint32_t *partial) +{ + struct scatterlist *sg; + uint32_t cumulative_partial, sg_len; + dma_addr_t sg_dma_addr; + + if (sgx->num_bytes == sgx->tot_bytes) + return 0; + + sg = sgx->cur_sg; + cumulative_partial = sgx->tot_partial; + + sg_dma_addr = sg_dma_address(sg); + sg_len = sg_dma_len(sg); + + sgx->dma_addr = sg_dma_addr + sgx->bytes_consumed; + + if ((cumulative_partial + (sg_len - sgx->bytes_consumed)) >= blk_sz) { + sgx->dma_len = (blk_sz - cumulative_partial); + sgx->tot_partial = 0; + sgx->num_bytes += blk_sz; + *partial = 0; + } else { + sgx->dma_len = sg_len - sgx->bytes_consumed; + sgx->tot_partial += sgx->dma_len; + *partial = 1; + } + + sgx->bytes_consumed += sgx->dma_len; + + if (sg_len == sgx->bytes_consumed) { + sg = sg_next(sg); + sgx->num_sg++; + sgx->cur_sg = sg; + sgx->bytes_consumed = 0; + } + + return 1; +} + +static int +qla24xx_walk_and_build_sglist_no_difb(struct qla_hw_data *ha, srb_t *sp, + uint32_t *dsd, uint16_t tot_dsds) +{ + void *next_dsd; + uint8_t avail_dsds = 0; + uint32_t dsd_list_len; + struct dsd_dma *dsd_ptr; + struct scatterlist *sg_prot; + uint32_t *cur_dsd = dsd; + uint16_t used_dsds = tot_dsds; + + uint32_t prot_int; + uint32_t partial; + struct qla2_sgx sgx; + dma_addr_t sle_dma; + uint32_t sle_dma_len, tot_prot_dma_len = 0; + struct scsi_cmnd *cmd = sp->cmd; + + prot_int = cmd->device->sector_size; + + memset(&sgx, 0, sizeof(struct qla2_sgx)); + sgx.tot_bytes = scsi_bufflen(sp->cmd); + sgx.cur_sg = scsi_sglist(sp->cmd); + sgx.sp = sp; + + sg_prot = scsi_prot_sglist(sp->cmd); + + while (qla24xx_get_one_block_sg(prot_int, &sgx, &partial)) { + + sle_dma = sgx.dma_addr; + sle_dma_len = sgx.dma_len; +alloc_and_fill: + /* Allocate additional continuation packets? */ + if (avail_dsds == 0) { + avail_dsds = (used_dsds > QLA_DSDS_PER_IOCB) ? + QLA_DSDS_PER_IOCB : used_dsds; + dsd_list_len = (avail_dsds + 1) * 12; + used_dsds -= avail_dsds; + + /* allocate tracking DS */ + dsd_ptr = kzalloc(sizeof(struct dsd_dma), GFP_ATOMIC); + if (!dsd_ptr) + return 1; + + /* allocate new list */ + dsd_ptr->dsd_addr = next_dsd = + dma_pool_alloc(ha->dl_dma_pool, GFP_ATOMIC, + &dsd_ptr->dsd_list_dma); + + if (!next_dsd) { + /* + * Need to cleanup only this dsd_ptr, rest + * will be done by sp_free_dma() + */ + kfree(dsd_ptr); + return 1; + } + + list_add_tail(&dsd_ptr->list, + &((struct crc_context *)sp->ctx)->dsd_list); + + sp->flags |= SRB_CRC_CTX_DSD_VALID; + + /* add new list to cmd iocb or last list */ + *cur_dsd++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma)); + *cur_dsd++ = cpu_to_le32(MSD(dsd_ptr->dsd_list_dma)); + *cur_dsd++ = dsd_list_len; + cur_dsd = (uint32_t *)next_dsd; + } + *cur_dsd++ = cpu_to_le32(LSD(sle_dma)); + *cur_dsd++ = cpu_to_le32(MSD(sle_dma)); + *cur_dsd++ = cpu_to_le32(sle_dma_len); + avail_dsds--; + + if (partial == 0) { + /* Got a full protection interval */ + sle_dma = sg_dma_address(sg_prot) + tot_prot_dma_len; + sle_dma_len = 8; + + tot_prot_dma_len += sle_dma_len; + if (tot_prot_dma_len == sg_dma_len(sg_prot)) { + tot_prot_dma_len = 0; + sg_prot = sg_next(sg_prot); + } + + partial = 1; /* So as to not re-enter this block */ + goto alloc_and_fill; + } + } + /* Null termination */ + *cur_dsd++ = 0; + *cur_dsd++ = 0; + *cur_dsd++ = 0; + return 0; +} static int qla24xx_walk_and_build_sglist(struct qla_hw_data *ha, srb_t *sp, uint32_t *dsd, uint16_t tot_dsds) @@ -981,7 +1140,7 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, struct scsi_cmnd *cmd; struct scatterlist *cur_seg; int sgc; - uint32_t total_bytes; + uint32_t total_bytes = 0; uint32_t data_bytes; uint32_t dif_bytes; uint8_t bundling = 1; @@ -1023,8 +1182,10 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, __constant_cpu_to_le16(CF_READ_DATA); } - tot_prot_dsds = scsi_prot_sg_count(cmd); - if (!tot_prot_dsds) + if ((scsi_get_prot_op(sp->cmd) == SCSI_PROT_READ_INSERT) || + (scsi_get_prot_op(sp->cmd) == SCSI_PROT_WRITE_STRIP) || + (scsi_get_prot_op(sp->cmd) == SCSI_PROT_READ_STRIP) || + (scsi_get_prot_op(sp->cmd) == SCSI_PROT_WRITE_INSERT)) bundling = 0; /* Allocate CRC context from global pool */ @@ -1107,15 +1268,28 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, cmd_pkt->fcp_rsp_dseg_len = 0; /* Let response come in status iocb */ /* Compute dif len and adjust data len to incude protection */ - total_bytes = data_bytes; dif_bytes = 0; blk_size = cmd->device->sector_size; - if (scsi_get_prot_op(cmd) != SCSI_PROT_NORMAL) { - dif_bytes = (data_bytes / blk_size) * 8; - total_bytes += dif_bytes; + dif_bytes = (data_bytes / blk_size) * 8; + + switch (scsi_get_prot_op(sp->cmd)) { + case SCSI_PROT_READ_INSERT: + case SCSI_PROT_WRITE_STRIP: + total_bytes = data_bytes; + data_bytes += dif_bytes; + break; + + case SCSI_PROT_READ_STRIP: + case SCSI_PROT_WRITE_INSERT: + case SCSI_PROT_READ_PASS: + case SCSI_PROT_WRITE_PASS: + total_bytes = data_bytes + dif_bytes; + break; + default: + BUG(); } - if (!ql2xenablehba_err_chk) + if (!qla2x00_hba_err_chk_enabled(scsi_get_prot_op(cmd))) fw_prot_opts |= 0x10; /* Disable Guard tag checking */ if (!bundling) { @@ -1151,7 +1325,12 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, cmd_pkt->control_flags |= __constant_cpu_to_le16(CF_DATA_SEG_DESCR_ENABLE); - if (qla24xx_walk_and_build_sglist(ha, sp, cur_dsd, + + if (!bundling && tot_prot_dsds) { + if (qla24xx_walk_and_build_sglist_no_difb(ha, sp, + cur_dsd, tot_dsds)) + goto crc_queuing_error; + } else if (qla24xx_walk_and_build_sglist(ha, sp, cur_dsd, (tot_dsds - tot_prot_dsds))) goto crc_queuing_error; @@ -1414,6 +1593,22 @@ qla24xx_dif_start_scsi(srb_t *sp) goto queuing_error; else sp->flags |= SRB_DMA_VALID; + + if ((scsi_get_prot_op(cmd) == SCSI_PROT_READ_INSERT) || + (scsi_get_prot_op(cmd) == SCSI_PROT_WRITE_STRIP)) { + struct qla2_sgx sgx; + uint32_t partial; + + memset(&sgx, 0, sizeof(struct qla2_sgx)); + sgx.tot_bytes = scsi_bufflen(cmd); + sgx.cur_sg = scsi_sglist(cmd); + sgx.sp = sp; + + nseg = 0; + while (qla24xx_get_one_block_sg( + cmd->device->sector_size, &sgx, &partial)) + nseg++; + } } else nseg = 0; @@ -1428,6 +1623,11 @@ qla24xx_dif_start_scsi(srb_t *sp) goto queuing_error; else sp->flags |= SRB_CRC_PROT_DMA_VALID; + + if ((scsi_get_prot_op(cmd) == SCSI_PROT_READ_INSERT) || + (scsi_get_prot_op(cmd) == SCSI_PROT_WRITE_STRIP)) { + nseg = scsi_bufflen(cmd) / cmd->device->sector_size; + } } else { nseg = 0; } @@ -1454,6 +1654,7 @@ qla24xx_dif_start_scsi(srb_t *sp) /* Build header part of command packet (excluding the OPCODE). */ req->current_outstanding_cmd = handle; req->outstanding_cmds[handle] = sp; + sp->handle = handle; sp->cmd->host_scribble = (unsigned char *)(unsigned long)handle; req->cnt -= req_cnt; diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c index b16b7725dee0..53339f10a598 100644 --- a/drivers/scsi/qla2xxx/qla_isr.c +++ b/drivers/scsi/qla2xxx/qla_isr.c @@ -1435,25 +1435,27 @@ struct scsi_dif_tuple { * ASC/ASCQ fields in the sense buffer with ILLEGAL_REQUEST * to indicate to the kernel that the HBA detected error. */ -static inline void +static inline int qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) { struct scsi_qla_host *vha = sp->fcport->vha; struct scsi_cmnd *cmd = sp->cmd; - struct scsi_dif_tuple *ep = - (struct scsi_dif_tuple *)&sts24->data[20]; - struct scsi_dif_tuple *ap = - (struct scsi_dif_tuple *)&sts24->data[12]; + uint8_t *ap = &sts24->data[12]; + uint8_t *ep = &sts24->data[20]; uint32_t e_ref_tag, a_ref_tag; uint16_t e_app_tag, a_app_tag; uint16_t e_guard, a_guard; - e_ref_tag = be32_to_cpu(ep->ref_tag); - a_ref_tag = be32_to_cpu(ap->ref_tag); - e_app_tag = be16_to_cpu(ep->app_tag); - a_app_tag = be16_to_cpu(ap->app_tag); - e_guard = be16_to_cpu(ep->guard); - a_guard = be16_to_cpu(ap->guard); + /* + * swab32 of the "data" field in the beginning of qla2x00_status_entry() + * would make guard field appear at offset 2 + */ + a_guard = le16_to_cpu(*(uint16_t *)(ap + 2)); + a_app_tag = le16_to_cpu(*(uint16_t *)(ap + 0)); + a_ref_tag = le32_to_cpu(*(uint32_t *)(ap + 4)); + e_guard = le16_to_cpu(*(uint16_t *)(ep + 2)); + e_app_tag = le16_to_cpu(*(uint16_t *)(ep + 0)); + e_ref_tag = le32_to_cpu(*(uint32_t *)(ep + 4)); ql_dbg(ql_dbg_io, vha, 0x3023, "iocb(s) %p Returned STATUS.\n", sts24); @@ -1465,6 +1467,63 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) cmd->cmnd[0], (u64)scsi_get_lba(cmd), a_ref_tag, e_ref_tag, a_app_tag, e_app_tag, a_guard, e_guard); + /* + * Ignore sector if: + * For type 3: ref & app tag is all 'f's + * For type 0,1,2: app tag is all 'f's + */ + if ((a_app_tag == 0xffff) && + ((scsi_get_prot_type(cmd) != SCSI_PROT_DIF_TYPE3) || + (a_ref_tag == 0xffffffff))) { + uint32_t blocks_done, resid; + sector_t lba_s = scsi_get_lba(cmd); + + /* 2TB boundary case covered automatically with this */ + blocks_done = e_ref_tag - (uint32_t)lba_s + 1; + + resid = scsi_bufflen(cmd) - (blocks_done * + cmd->device->sector_size); + + scsi_set_resid(cmd, resid); + cmd->result = DID_OK << 16; + + /* Update protection tag */ + if (scsi_prot_sg_count(cmd)) { + uint32_t i, j = 0, k = 0, num_ent; + struct scatterlist *sg; + struct sd_dif_tuple *spt; + + /* Patch the corresponding protection tags */ + scsi_for_each_prot_sg(cmd, sg, + scsi_prot_sg_count(cmd), i) { + num_ent = sg_dma_len(sg) / 8; + if (k + num_ent < blocks_done) { + k += num_ent; + continue; + } + j = blocks_done - k - 1; + k = blocks_done; + break; + } + + if (k != blocks_done) { + qla_printk(KERN_WARNING, sp->fcport->vha->hw, + "unexpected tag values tag:lba=%x:%lx)\n", + e_ref_tag, lba_s); + return 1; + } + + spt = page_address(sg_page(sg)) + sg->offset; + spt += j; + + spt->app_tag = 0xffff; + if (scsi_get_prot_type(cmd) == SCSI_PROT_DIF_TYPE3) + spt->ref_tag = 0xffffffff; + } + + return 0; + } + /* check guard */ if (e_guard != a_guard) { scsi_build_sense_buffer(1, cmd->sense_buffer, ILLEGAL_REQUEST, @@ -1472,7 +1531,7 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) set_driver_byte(cmd, DRIVER_SENSE); set_host_byte(cmd, DID_ABORT); cmd->result |= SAM_STAT_CHECK_CONDITION << 1; - return; + return 1; } /* check appl tag */ @@ -1482,7 +1541,7 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) set_driver_byte(cmd, DRIVER_SENSE); set_host_byte(cmd, DID_ABORT); cmd->result |= SAM_STAT_CHECK_CONDITION << 1; - return; + return 1; } /* check ref tag */ @@ -1492,8 +1551,9 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) set_driver_byte(cmd, DRIVER_SENSE); set_host_byte(cmd, DID_ABORT); cmd->result |= SAM_STAT_CHECK_CONDITION << 1; - return; + return 1; } + return 1; } /** @@ -1767,7 +1827,7 @@ check_scsi_status: break; case CS_DIF_ERROR: - qla2x00_handle_dif_error(sp, sts24); + logit = qla2x00_handle_dif_error(sp, sts24); break; default: cp->result = DID_ERROR << 16; diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c index e02df276804e..d65a3005b439 100644 --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -106,17 +106,21 @@ MODULE_PARM_DESC(ql2xmaxqdepth, "Maximum queue depth to report for target devices."); /* Do not change the value of this after module load */ -int ql2xenabledif = 1; +int ql2xenabledif = 0; module_param(ql2xenabledif, int, S_IRUGO|S_IWUSR); MODULE_PARM_DESC(ql2xenabledif, " Enable T10-CRC-DIF " - " Default is 0 - No DIF Support. 1 - Enable it"); + " Default is 0 - No DIF Support. 1 - Enable it" + ", 2 - Enable DIF for all types, except Type 0."); -int ql2xenablehba_err_chk; +int ql2xenablehba_err_chk = 2; module_param(ql2xenablehba_err_chk, int, S_IRUGO|S_IWUSR); MODULE_PARM_DESC(ql2xenablehba_err_chk, - " Enable T10-CRC-DIF Error isolation by HBA" - " Default is 0 - Error isolation disabled, 1 - Enable it"); + " Enable T10-CRC-DIF Error isolation by HBA:\n" + " Default is 1.\n" + " 0 -- Error isolation disabled\n" + " 1 -- Error isolation enabled only for DIX Type 0\n" + " 2 -- Error isolation enabled for all Types\n"); int ql2xiidmaenable=1; module_param(ql2xiidmaenable, int, S_IRUGO); @@ -2380,11 +2384,14 @@ skip_dpc: if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) { if (ha->fw_attributes & BIT_4) { + int prot = 0; base_vha->flags.difdix_supported = 1; ql_dbg(ql_dbg_init, base_vha, 0x00f1, "Registering for DIF/DIX type 1 and 3 protection.\n"); + if (ql2xenabledif == 1) + prot = SHOST_DIX_TYPE0_PROTECTION; scsi_host_set_prot(host, - SHOST_DIF_TYPE1_PROTECTION + prot | SHOST_DIF_TYPE1_PROTECTION | SHOST_DIF_TYPE2_PROTECTION | SHOST_DIF_TYPE3_PROTECTION | SHOST_DIX_TYPE1_PROTECTION From e02587d777bfb398f70709fd3a92fa0154959003 Mon Sep 17 00:00:00 2001 From: Arun Easi Date: Tue, 16 Aug 2011 11:29:23 -0700 Subject: [PATCH 13/25] [SCSI] qla2xxx: T10 DIF - Fix incorrect error reporting. This fix: - Disables app tag peeking; correct tag check will be added when the SCSI API is available. - Always derive ref_tag from scsi_get_lba() - Removes incorrect swap of FCP_LUN in FCP_CMND - Moves app-tag error check before ref-tag check. The reason being, currently there is no interface in SCSI to retrieve the app-tag for protection I/Os, so driver puts zero for app-tag in the firmware interface, but requests not to validate it, but when a ref-tag error is detected by firmware, it would put expected/actual tags for all the protection tags (guard/app/ref). As driver checks for app tag error first, a ref-tag error is incorrectly flagged as app-tag error. - Convert HBA specific checks to capability based. Signed-off-by: Arun Easi Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_attr.c | 2 +- drivers/scsi/qla2xxx/qla_dbg.c | 36 ++++++++++---------- drivers/scsi/qla2xxx/qla_def.h | 2 ++ drivers/scsi/qla2xxx/qla_inline.h | 12 +++++-- drivers/scsi/qla2xxx/qla_iocb.c | 55 ++++++++++++------------------- drivers/scsi/qla2xxx/qla_isr.c | 19 ++++++----- drivers/scsi/qla2xxx/qla_mid.c | 2 +- drivers/scsi/qla2xxx/qla_os.c | 4 +-- 8 files changed, 65 insertions(+), 67 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c index 810067099801..a31e05f3bfd4 100644 --- a/drivers/scsi/qla2xxx/qla_attr.c +++ b/drivers/scsi/qla2xxx/qla_attr.c @@ -1786,7 +1786,7 @@ qla24xx_vport_create(struct fc_vport *fc_vport, bool disable) fc_vport_set_state(fc_vport, FC_VPORT_LINKDOWN); } - if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) { + if (IS_T10_PI_CAPABLE(ha) && ql2xenabledif) { if (ha->fw_attributes & BIT_4) { int prot = 0; vha->flags.difdix_supported = 1; diff --git a/drivers/scsi/qla2xxx/qla_dbg.c b/drivers/scsi/qla2xxx/qla_dbg.c index 2155071f3100..d79cd8a5f831 100644 --- a/drivers/scsi/qla2xxx/qla_dbg.c +++ b/drivers/scsi/qla2xxx/qla_dbg.c @@ -8,24 +8,24 @@ /* * Table for showing the current message id in use for particular level * Change this table for addition of log/debug messages. - * ----------------------------------------------------- - * | Level | Last Value Used | - * ----------------------------------------------------- - * | Module Init and Probe | 0x0116 | - * | Mailbox commands | 0x111e | - * | Device Discovery | 0x2083 | - * | Queue Command and IO tracing | 0x302e | - * | DPC Thread | 0x401c | - * | Async Events | 0x5059 | - * | Timer Routines | 0x600d | - * | User Space Interactions | 0x709c | - * | Task Management | 0x8043 | - * | AER/EEH | 0x900f | - * | Virtual Port | 0xa007 | - * | ISP82XX Specific | 0xb027 | - * | MultiQ | 0xc00b | - * | Misc | 0xd00b | - * ----------------------------------------------------- + * ---------------------------------------------------------------------- + * | Level | Last Value Used | Holes | + * ---------------------------------------------------------------------- + * | Module Init and Probe | 0x0116 | | + * | Mailbox commands | 0x1126 | | + * | Device Discovery | 0x2083 | | + * | Queue Command and IO tracing | 0x302e | 0x3008 | + * | DPC Thread | 0x401c | | + * | Async Events | 0x5059 | | + * | Timer Routines | 0x600d | | + * | User Space Interactions | 0x709d | | + * | Task Management | 0x8041 | | + * | AER/EEH | 0x900f | | + * | Virtual Port | 0xa007 | | + * | ISP82XX Specific | 0xb04f | | + * | MultiQ | 0xc00b | | + * | Misc | 0xd00b | | + * ---------------------------------------------------------------------- */ #include "qla_def.h" diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h index cc5a79259d33..a03eaf40f377 100644 --- a/drivers/scsi/qla2xxx/qla_def.h +++ b/drivers/scsi/qla2xxx/qla_def.h @@ -2529,6 +2529,7 @@ struct qla_hw_data { #define DT_ISP8021 BIT_14 #define DT_ISP_LAST (DT_ISP8021 << 1) +#define DT_T10_PI BIT_25 #define DT_IIDMA BIT_26 #define DT_FWI2 BIT_27 #define DT_ZIO_SUPPORTED BIT_28 @@ -2572,6 +2573,7 @@ struct qla_hw_data { #define IS_NOCACHE_VPD_TYPE(ha) (IS_QLA81XX(ha)) #define IS_ALOGIO_CAPABLE(ha) (IS_QLA23XX(ha) || IS_FWI2_CAPABLE(ha)) +#define IS_T10_PI_CAPABLE(ha) ((ha)->device_type & DT_T10_PI) #define IS_IIDMA_CAPABLE(ha) ((ha)->device_type & DT_IIDMA) #define IS_FWI2_CAPABLE(ha) ((ha)->device_type & DT_FWI2) #define IS_ZIO_SUPPORTED(ha) ((ha)->device_type & DT_ZIO_SUPPORTED) diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h index c06e5f9b431e..9902834e0b74 100644 --- a/drivers/scsi/qla2xxx/qla_inline.h +++ b/drivers/scsi/qla2xxx/qla_inline.h @@ -104,9 +104,17 @@ qla2x00_set_fcport_state(fc_port_t *fcport, int state) } static inline int -qla2x00_hba_err_chk_enabled(unsigned char op) +qla2x00_hba_err_chk_enabled(srb_t *sp) { - switch (op) { + /* + * Uncomment when corresponding SCSI changes are done. + * + if (!sp->cmd->prot_chk) + return 0; + * + */ + + switch (scsi_get_prot_op(sp->cmd)) { case SCSI_PROT_READ_STRIP: case SCSI_PROT_WRITE_INSERT: if (ql2xenablehba_err_chk >= 1) diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c index 09ad3ce60064..dbec89622a0f 100644 --- a/drivers/scsi/qla2xxx/qla_iocb.c +++ b/drivers/scsi/qla2xxx/qla_iocb.c @@ -709,12 +709,11 @@ struct fw_dif_context { * */ static inline void -qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, +qla24xx_set_t10dif_tags(srb_t *sp, struct fw_dif_context *pkt, unsigned int protcnt) { - struct sd_dif_tuple *spt; + struct scsi_cmnd *cmd = sp->cmd; scsi_qla_host_t *vha = shost_priv(cmd->device->host); - unsigned char op = scsi_get_prot_op(cmd); switch (scsi_get_prot_type(cmd)) { case SCSI_PROT_DIF_TYPE0: @@ -724,6 +723,10 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, */ pkt->ref_tag = cpu_to_le32((uint32_t) (0xffffffff & scsi_get_lba(cmd))); + + if (!qla2x00_hba_err_chk_enabled(sp)) + break; + pkt->ref_tag_mask[0] = 0xff; pkt->ref_tag_mask[1] = 0xff; pkt->ref_tag_mask[2] = 0xff; @@ -735,20 +738,16 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, * match LBA in CDB + N */ case SCSI_PROT_DIF_TYPE2: - if (!qla2x00_hba_err_chk_enabled(op)) - break; - - if (scsi_prot_sg_count(cmd)) { - spt = page_address(sg_page(scsi_prot_sglist(cmd))) + - scsi_prot_sglist(cmd)[0].offset; - pkt->app_tag = swab32(spt->app_tag); - pkt->app_tag_mask[0] = 0xff; - pkt->app_tag_mask[1] = 0xff; - } + pkt->app_tag = __constant_cpu_to_le16(0); + pkt->app_tag_mask[0] = 0x0; + pkt->app_tag_mask[1] = 0x0; pkt->ref_tag = cpu_to_le32((uint32_t) (0xffffffff & scsi_get_lba(cmd))); + if (!qla2x00_hba_err_chk_enabled(sp)) + break; + /* enable ALL bytes of the ref tag */ pkt->ref_tag_mask[0] = 0xff; pkt->ref_tag_mask[1] = 0xff; @@ -768,26 +767,15 @@ qla24xx_set_t10dif_tags(struct scsi_cmnd *cmd, struct fw_dif_context *pkt, * 16 bit app tag. */ case SCSI_PROT_DIF_TYPE1: - if (!qla2x00_hba_err_chk_enabled(op)) + pkt->ref_tag = cpu_to_le32((uint32_t) + (0xffffffff & scsi_get_lba(cmd))); + pkt->app_tag = __constant_cpu_to_le16(0); + pkt->app_tag_mask[0] = 0x0; + pkt->app_tag_mask[1] = 0x0; + + if (!qla2x00_hba_err_chk_enabled(sp)) break; - if (protcnt && (op == SCSI_PROT_WRITE_STRIP || - op == SCSI_PROT_WRITE_PASS)) { - spt = page_address(sg_page(scsi_prot_sglist(cmd))) + - scsi_prot_sglist(cmd)[0].offset; - ql_dbg(ql_dbg_io, vha, 0x3008, - "LBA from user %p, lba = 0x%x for cmd=%p.\n", - spt, (int)spt->ref_tag, cmd); - pkt->ref_tag = swab32(spt->ref_tag); - pkt->app_tag_mask[0] = 0x0; - pkt->app_tag_mask[1] = 0x0; - } else { - pkt->ref_tag = cpu_to_le32((uint32_t) - (0xffffffff & scsi_get_lba(cmd))); - pkt->app_tag = __constant_cpu_to_le16(0); - pkt->app_tag_mask[0] = 0x0; - pkt->app_tag_mask[1] = 0x0; - } /* enable ALL bytes of the ref tag */ pkt->ref_tag_mask[0] = 0xff; pkt->ref_tag_mask[1] = 0xff; @@ -1208,7 +1196,7 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, INIT_LIST_HEAD(&crc_ctx_pkt->dsd_list); - qla24xx_set_t10dif_tags(cmd, (struct fw_dif_context *) + qla24xx_set_t10dif_tags(sp, (struct fw_dif_context *) &crc_ctx_pkt->ref_tag, tot_prot_dsds); cmd_pkt->crc_context_address[0] = cpu_to_le32(LSD(crc_ctx_dma)); @@ -1237,7 +1225,6 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, fcp_cmnd->additional_cdb_len |= 2; int_to_scsilun(sp->cmd->device->lun, &fcp_cmnd->lun); - host_to_fcp_swap((uint8_t *)&fcp_cmnd->lun, sizeof(fcp_cmnd->lun)); memcpy(fcp_cmnd->cdb, cmd->cmnd, cmd->cmd_len); cmd_pkt->fcp_cmnd_dseg_len = cpu_to_le16(fcp_cmnd_len); cmd_pkt->fcp_cmnd_dseg_address[0] = cpu_to_le32( @@ -1289,7 +1276,7 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt, BUG(); } - if (!qla2x00_hba_err_chk_enabled(scsi_get_prot_op(cmd))) + if (!qla2x00_hba_err_chk_enabled(sp)) fw_prot_opts |= 0x10; /* Disable Guard tag checking */ if (!bundling) { diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c index 53339f10a598..ec53e87781a5 100644 --- a/drivers/scsi/qla2xxx/qla_isr.c +++ b/drivers/scsi/qla2xxx/qla_isr.c @@ -1534,6 +1534,16 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) return 1; } + /* check ref tag */ + if (e_ref_tag != a_ref_tag) { + scsi_build_sense_buffer(1, cmd->sense_buffer, ILLEGAL_REQUEST, + 0x10, 0x3); + set_driver_byte(cmd, DRIVER_SENSE); + set_host_byte(cmd, DID_ABORT); + cmd->result |= SAM_STAT_CHECK_CONDITION << 1; + return 1; + } + /* check appl tag */ if (e_app_tag != a_app_tag) { scsi_build_sense_buffer(1, cmd->sense_buffer, ILLEGAL_REQUEST, @@ -1544,15 +1554,6 @@ qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24) return 1; } - /* check ref tag */ - if (e_ref_tag != a_ref_tag) { - scsi_build_sense_buffer(1, cmd->sense_buffer, ILLEGAL_REQUEST, - 0x10, 0x3); - set_driver_byte(cmd, DRIVER_SENSE); - set_host_byte(cmd, DID_ABORT); - cmd->result |= SAM_STAT_CHECK_CONDITION << 1; - return 1; - } return 1; } diff --git a/drivers/scsi/qla2xxx/qla_mid.c b/drivers/scsi/qla2xxx/qla_mid.c index c706ed370000..f488cc69fc79 100644 --- a/drivers/scsi/qla2xxx/qla_mid.c +++ b/drivers/scsi/qla2xxx/qla_mid.c @@ -472,7 +472,7 @@ qla24xx_create_vhost(struct fc_vport *fc_vport) host->can_queue = base_vha->req->length + 128; host->this_id = 255; host->cmd_per_lun = 3; - if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) + if (IS_T10_PI_CAPABLE(ha) && ql2xenabledif) host->max_cmd_len = 32; else host->max_cmd_len = MAX_CMDSZ; diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c index d65a3005b439..f57c292845a5 100644 --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -2255,7 +2255,7 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id) host->this_id = 255; host->cmd_per_lun = 3; host->unique_id = host->host_no; - if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) + if (IS_T10_PI_CAPABLE(ha) && ql2xenabledif) host->max_cmd_len = 32; else host->max_cmd_len = MAX_CMDSZ; @@ -2382,7 +2382,7 @@ skip_dpc: "Detected hba at address=%p.\n", ha); - if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) { + if (IS_T10_PI_CAPABLE(ha) && ql2xenabledif) { if (ha->fw_attributes & BIT_4) { int prot = 0; base_vha->flags.difdix_supported = 1; From 42cd4f5dc2a3de31bfd24642ab4e8b21834a6b78 Mon Sep 17 00:00:00 2001 From: Chad Dupuis Date: Tue, 16 Aug 2011 11:29:24 -0700 Subject: [PATCH 14/25] [SCSI] qla2xxx: Fix qla24xx revision check while enabling interrupts. Since we enable interrupts before initializing the firmware, use the chip revision from PCI config space directly to perform the chip revision check. Also remove the unnecessary firmware attributes test. Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_isr.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c index ec53e87781a5..477767fcfd1e 100644 --- a/drivers/scsi/qla2xxx/qla_isr.c +++ b/drivers/scsi/qla2xxx/qla_isr.c @@ -2529,11 +2529,10 @@ qla2x00_request_irqs(struct qla_hw_data *ha, struct rsp_que *rsp) goto skip_msi; } - if (IS_QLA2432(ha) && (ha->pdev->revision < QLA_MSIX_CHIP_REV_24XX || - !QLA_MSIX_FW_MODE_1(ha->fw_attributes))) { + if (IS_QLA2432(ha) && (ha->pdev->revision < QLA_MSIX_CHIP_REV_24XX)) { ql_log(ql_log_warn, vha, 0x0035, "MSI-X; Unsupported ISP2432 (0x%X, 0x%X).\n", - ha->pdev->revision, ha->fw_attributes); + ha->pdev->revision, QLA_MSIX_CHIP_REV_24XX); goto skip_msix; } From 7594206493880007fd68a18d6e9f380a1afe20d4 Mon Sep 17 00:00:00 2001 From: Saurav Kashyap Date: Tue, 16 Aug 2011 11:29:25 -0700 Subject: [PATCH 15/25] [SCSI] qla2xxx: Acquire hardware lock while manipulating dsd list. The dsd list shouldn't be manipulated without taking the per host hardware lock to prevent multiple callers from trampling upon one another. Signed-off-by: Saurav Kashyap Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_os.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c index f57c292845a5..2caab83c4c9f 100644 --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -913,7 +913,10 @@ qla2xxx_eh_abort(struct scsi_cmnd *cmd) "Abort command mbx success.\n"); wait = 1; } + + spin_lock_irqsave(&ha->hardware_lock, flags); qla2x00_sp_compl(ha, sp); + spin_unlock_irqrestore(&ha->hardware_lock, flags); /* Wait for the command to be returned. */ if (wait) { From bc91ade9b7bc274d625c9b24c04d365a2daf481e Mon Sep 17 00:00:00 2001 From: Chad Dupuis Date: Tue, 16 Aug 2011 11:29:26 -0700 Subject: [PATCH 16/25] [SCSI] qla2xxx: Double check for command completion if abort mailbox command fails. Close a small window where we could falsely fail an abort request if the mailbox command fails but the command was returned during interrupt context. Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_os.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c index 2caab83c4c9f..4cace3f20c04 100644 --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -918,6 +918,10 @@ qla2xxx_eh_abort(struct scsi_cmnd *cmd) qla2x00_sp_compl(ha, sp); spin_unlock_irqrestore(&ha->hardware_lock, flags); + /* Did the command return during mailbox execution? */ + if (ret == FAILED && !CMD_SP(cmd)) + ret = SUCCESS; + /* Wait for the command to be returned. */ if (wait) { if (qla2x00_eh_wait_on_command(cmd) != QLA_SUCCESS) { From 3553d343e7acc418988cb8f22cd5b4976e7b484a Mon Sep 17 00:00:00 2001 From: Saurav Kashyap Date: Tue, 16 Aug 2011 11:29:27 -0700 Subject: [PATCH 17/25] [SCSI] qla2xxx: Save and restore irq in the response queue interrupt handler. Signed-off-by: Saurav Kashyap Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_nx.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_nx.c b/drivers/scsi/qla2xxx/qla_nx.c index 5cbf33a50b14..02704fe8afab 100644 --- a/drivers/scsi/qla2xxx/qla_nx.c +++ b/drivers/scsi/qla2xxx/qla_nx.c @@ -2208,6 +2208,7 @@ qla82xx_msix_rsp_q(int irq, void *dev_id) struct qla_hw_data *ha; struct rsp_que *rsp; struct device_reg_82xx __iomem *reg; + unsigned long flags; rsp = (struct rsp_que *) dev_id; if (!rsp) { @@ -2218,11 +2219,11 @@ qla82xx_msix_rsp_q(int irq, void *dev_id) ha = rsp->hw; reg = &ha->iobase->isp82; - spin_lock_irq(&ha->hardware_lock); + spin_lock_irqsave(&ha->hardware_lock, flags); vha = pci_get_drvdata(ha->pdev); qla24xx_process_response_queue(vha, rsp); WRT_REG_DWORD(®->host_int, 0); - spin_unlock_irq(&ha->hardware_lock); + spin_unlock_irqrestore(&ha->hardware_lock, flags); return IRQ_HANDLED; } From 58b48576966ed0afd3f63ef17480ec12748a7119 Mon Sep 17 00:00:00 2001 From: Andrew Vasquez Date: Tue, 16 Aug 2011 11:29:28 -0700 Subject: [PATCH 18/25] [SCSI] qla2xxx: Correct inadvertent loop state transitions during port-update handling. Transitioning to a LOOP_UPDATE loop-state could cause the driver to miss normal link/target processing. LOOP_UPDATE is a crufty artifact leftover from at time the driver performed it's own internal command-queuing. Safely remove this state. Signed-off-by: Andrew Vasquez Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_init.c | 3 --- drivers/scsi/qla2xxx/qla_isr.c | 1 - 2 files changed, 4 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c index def694271bf7..37da04d3db26 100644 --- a/drivers/scsi/qla2xxx/qla_init.c +++ b/drivers/scsi/qla2xxx/qla_init.c @@ -3838,15 +3838,12 @@ qla2x00_loop_resync(scsi_qla_host_t *vha) req = vha->req; rsp = req->rsp; - atomic_set(&vha->loop_state, LOOP_UPDATE); clear_bit(ISP_ABORT_RETRY, &vha->dpc_flags); if (vha->flags.online) { if (!(rval = qla2x00_fw_ready(vha))) { /* Wait at most MAX_TARGET RSCNs for a stable link. */ wait_time = 256; do { - atomic_set(&vha->loop_state, LOOP_UPDATE); - /* Issue a marker after FW becomes ready. */ qla2x00_marker(vha, req, rsp, 0, 0, MK_SYNC_ALL); diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c index 477767fcfd1e..646fc5263d50 100644 --- a/drivers/scsi/qla2xxx/qla_isr.c +++ b/drivers/scsi/qla2xxx/qla_isr.c @@ -719,7 +719,6 @@ skip_rio: vha->flags.rscn_queue_overflow = 1; } - atomic_set(&vha->loop_state, LOOP_UPDATE); atomic_set(&vha->loop_down_timer, 0); vha->flags.management_server_logged_in = 0; From 51cc9a8e5f610a0d0881b45410c37890e02a2f76 Mon Sep 17 00:00:00 2001 From: Saurav Kashyap Date: Tue, 16 Aug 2011 11:29:29 -0700 Subject: [PATCH 19/25] [SCSI] qla2xxx: Set the task attributes after memsetting fcp cmnd. The memset of the fcp_cmnd struct needs to be moved so that it will not zero-out valid data. Signed-off-by: Saurav Kashyap Signed-off-by: Chad Dupuis Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_nx.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/drivers/scsi/qla2xxx/qla_nx.c b/drivers/scsi/qla2xxx/qla_nx.c index 02704fe8afab..049807cda419 100644 --- a/drivers/scsi/qla2xxx/qla_nx.c +++ b/drivers/scsi/qla2xxx/qla_nx.c @@ -2839,6 +2839,16 @@ sufficient_dsds: int_to_scsilun(sp->cmd->device->lun, &cmd_pkt->lun); host_to_fcp_swap((uint8_t *)&cmd_pkt->lun, sizeof(cmd_pkt->lun)); + /* build FCP_CMND IU */ + memset(ctx->fcp_cmnd, 0, sizeof(struct fcp_cmnd)); + int_to_scsilun(sp->cmd->device->lun, &ctx->fcp_cmnd->lun); + ctx->fcp_cmnd->additional_cdb_len = additional_cdb_len; + + if (cmd->sc_data_direction == DMA_TO_DEVICE) + ctx->fcp_cmnd->additional_cdb_len |= 1; + else if (cmd->sc_data_direction == DMA_FROM_DEVICE) + ctx->fcp_cmnd->additional_cdb_len |= 2; + /* * Update tagged queuing modifier -- default is TSK_SIMPLE (0). */ @@ -2855,16 +2865,6 @@ sufficient_dsds: } } - /* build FCP_CMND IU */ - memset(ctx->fcp_cmnd, 0, sizeof(struct fcp_cmnd)); - int_to_scsilun(sp->cmd->device->lun, &ctx->fcp_cmnd->lun); - ctx->fcp_cmnd->additional_cdb_len = additional_cdb_len; - - if (cmd->sc_data_direction == DMA_TO_DEVICE) - ctx->fcp_cmnd->additional_cdb_len |= 1; - else if (cmd->sc_data_direction == DMA_FROM_DEVICE) - ctx->fcp_cmnd->additional_cdb_len |= 2; - memcpy(ctx->fcp_cmnd->cdb, cmd->cmnd, cmd->cmd_len); fcp_dl = (uint32_t *)(ctx->fcp_cmnd->cdb + 16 + From 7ca3c803e85080afdff4097e60fefec865027809 Mon Sep 17 00:00:00 2001 From: Chad Dupuis Date: Tue, 16 Aug 2011 11:29:30 -0700 Subject: [PATCH 20/25] [SCSI] qla2xxx: Update version number to 8.03.07.07-k. Signed-off-by: James Bottomley --- drivers/scsi/qla2xxx/qla_version.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/qla2xxx/qla_version.h b/drivers/scsi/qla2xxx/qla_version.h index 062c97bf62f5..13b6357c1fa2 100644 --- a/drivers/scsi/qla2xxx/qla_version.h +++ b/drivers/scsi/qla2xxx/qla_version.h @@ -7,7 +7,7 @@ /* * Driver version */ -#define QLA2XXX_VERSION "8.03.07.03-k" +#define QLA2XXX_VERSION "8.03.07.07-k" #define QLA_DRIVER_MAJOR_VER 8 #define QLA_DRIVER_MINOR_VER 3 From 848e7d5b46b9b0ee613a106bc460acf6a09a8546 Mon Sep 17 00:00:00 2001 From: Robert Love Date: Thu, 25 Aug 2011 12:40:47 -0700 Subject: [PATCH 21/25] [SCSI] fcoe: Fix deadlock between fip's recv_work and rtnl The rtnl cannot be held durrng the fcoe_interface_put. If it is the last reference on the fcoe_interface the fcoe_ctlr_destroy will be called as a part of the cleanup, ultimately calling cancel_work_sync(&fip->recv_work); If we are processing a flogi response we will be in the recv_work context and we will lock the rtnl to add a new unicast MAC address. This is how the deadlock can occur. The fix is simply to move the rtnl_lock/unlock into fcoe_interface_cleanup so that it can be unlocked before fcoe_interface_put is called. Here is the lockdep report: Jul 21 11:26:35 bubba [ 223.870702] ul 21 11:26:35 bubba [ 223.870704] ======================================================= Jul 21 11:26:35 bubba [ 223.871255] [ INFO: possible circular locking dependency detected ] Jul 21 11:26:35 bubba [ 223.871530] 3.0.0-rc7+ #1 Jul 21 11:26:35 bubba [ 223.871797] ------------------------------------------------------- Jul 21 11:26:35 bubba [ 223.872072] lockdeptest.sh/3464 is trying to acquire lock: Jul 21 11:26:35 bubba [ 223.872345] ((&fip->recv_work) Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] wait_on_work+0x0/0xbd Jul 21 11:26:35 bubba [ 223.873022] Jul 21 11:26:35 bubba [ 223.873023] but task is already holding lock: Jul 21 11:26:35 bubba [ 223.873555] (rtnl_mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] rtnl_lock+0x12/0x14 Jul 21 11:26:35 bubba [ 223.874229] Jul 21 11:26:35 bubba [ 223.874230] which lock already depends on the new lock. Jul 21 11:26:35 bubba [ 223.874231] Jul 21 11:26:35 bubba [ 223.875032] Jul 21 11:26:35 bubba [ 223.875033] the existing dependency chain (in reverse order) is: Jul 21 11:26:35 bubba [ 223.875573] Jul 21 11:26:35 bubba [ 223.875573] -> #1 Jul 21 11:26:35 bubba (rtnl_mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba : Jul 21 11:26:35 bubba [ 223.876301] Jul 21 11:26:35 bubba [] lock_acquire+0xd2/0xf7 Jul 21 11:26:35 bubba [ 223.876645] Jul 21 11:26:35 bubba [] __mutex_lock_common+0x47/0x30d Jul 21 11:26:35 bubba [ 223.876991] Jul 21 11:26:35 bubba [] mutex_lock_nested+0x3b/0x40 Jul 21 11:26:35 bubba [ 223.877334] Jul 21 11:26:35 bubba [] rtnl_lock+0x12/0x14 Jul 21 11:26:35 bubba [ 223.877675] Jul 21 11:26:35 bubba [] fcoe_update_src_mac+0x2b/0x80 [fcoe] Jul 21 11:26:35 bubba [ 223.878022] Jul 21 11:26:35 bubba [] fcoe_flogi_resp+0x5e/0x79 [fcoe] Jul 21 11:26:35 bubba [ 223.878366] Jul 21 11:26:35 bubba [] fc_exch_recv+0x7f5/0x9da [libfc] Jul 21 11:26:35 bubba [ 223.878713] Jul 21 11:26:35 bubba [] fcoe_ctlr_recv_work+0x71f/0x10dc [libfcoe] Jul 21 11:26:35 bubba [ 223.879258] Jul 21 11:26:35 bubba [] process_one_work+0x1d7/0x347 Jul 21 11:26:35 bubba [ 223.879601] Jul 21 11:26:35 bubba [] worker_thread+0xf8/0x17c Jul 21 11:26:35 bubba [ 223.879944] Jul 21 11:26:35 bubba [] kthread+0x7d/0x85 Jul 21 11:26:35 bubba [ 223.880287] Jul 21 11:26:35 bubba [] kernel_thread_helper+0x4/0x10 Jul 21 11:26:35 bubba [ 223.880634] Jul 21 11:26:35 bubba [ 223.880635] -> #0 Jul 21 11:26:35 bubba ((&fip->recv_work) Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba : Jul 21 11:26:35 bubba [ 223.881357] Jul 21 11:26:35 bubba [] __lock_acquire+0xb1d/0xe2c Jul 21 11:26:35 bubba [ 223.881695] Jul 21 11:26:35 bubba [] lock_acquire+0xd2/0xf7 Jul 21 11:26:35 bubba [ 223.882033] Jul 21 11:26:35 bubba [] wait_on_work+0x50/0xbd Jul 21 11:26:35 bubba [ 223.882378] Jul 21 11:26:35 bubba [] __cancel_work_timer+0xb6/0xf4 Jul 21 11:26:35 bubba [ 223.882718] Jul 21 11:26:35 bubba [] cancel_work_sync+0xb/0xd Jul 21 11:26:35 bubba [ 223.883057] Jul 21 11:26:35 bubba [] fcoe_ctlr_destroy+0x1d/0x67 [libfcoe] Jul 21 11:26:35 bubba [ 223.883399] Jul 21 11:26:35 bubba [] fcoe_interface_release+0x21/0x45 [fcoe] Jul 21 11:26:35 bubba [ 223.883940] Jul 21 11:26:35 bubba [] kref_put+0x43/0x4d Jul 21 11:26:35 bubba [ 223.884280] Jul 21 11:26:35 bubba [] fcoe_interface_put+0x17/0x19 [fcoe] Jul 21 11:26:35 bubba [ 223.884624] Jul 21 11:26:35 bubba [] fcoe_interface_cleanup+0x188/0x193 [fcoe] Jul 21 11:26:35 bubba [ 223.885163] Jul 21 11:26:35 bubba [] fcoe_destroy+0x52/0x72 [fcoe] Jul 21 11:26:35 bubba [ 223.885502] Jul 21 11:26:35 bubba [] fcoe_transport_destroy+0xab/0x110 [libfcoe] Jul 21 11:26:35 bubba [ 223.886045] Jul 21 11:26:35 bubba [] param_attr_store+0x43/0x62 Jul 21 11:26:35 bubba [ 223.886385] Jul 21 11:26:35 bubba [] module_attr_store+0x21/0x25 Jul 21 11:26:35 bubba [ 223.886728] Jul 21 11:26:35 bubba [] sysfs_write_file+0x103/0x13f Jul 21 11:26:35 bubba [ 223.887068] Jul 21 11:26:35 bubba [] vfs_write+0xa7/0xfa Jul 21 11:26:35 bubba [ 223.887406] Jul 21 11:26:35 bubba [] sys_write+0x45/0x69 Jul 21 11:26:35 bubba [ 223.887742] Jul 21 11:26:35 bubba [] system_call_fastpath+0x16/0x1b Jul 21 11:26:35 bubba [ 223.888083] Jul 21 11:26:35 bubba [ 223.888084] other info that might help us debug this: Jul 21 11:26:35 bubba [ 223.888085] Jul 21 11:26:35 bubba [ 223.888879] Possible unsafe locking scenario: Jul 21 11:26:35 bubba [ 223.888881] Jul 21 11:26:35 bubba [ 223.889411] CPU0 CPU1 Jul 21 11:26:35 bubba [ 223.889683] ---- ---- Jul 21 11:26:35 bubba [ 223.889955] lock( Jul 21 11:26:35 bubba rtnl_mutex Jul 21 11:26:35 bubba ); Jul 21 11:26:35 bubba [ 223.890349] lock( Jul 21 11:26:35 bubba (&fip->recv_work) Jul 21 11:26:35 bubba ); Jul 21 11:26:35 bubba [ 223.890751] lock( Jul 21 11:26:35 bubba rtnl_mutex Jul 21 11:26:35 bubba ); Jul 21 11:26:35 bubba [ 223.891154] lock( Jul 21 11:26:35 bubba (&fip->recv_work) Jul 21 11:26:35 bubba ); Jul 21 11:26:35 bubba [ 223.891549] Jul 21 11:26:35 bubba [ 223.891550] *** DEADLOCK *** Jul 21 11:26:35 bubba [ 223.891551] Jul 21 11:26:35 bubba [ 223.892347] 6 locks held by lockdeptest.sh/3464: Jul 21 11:26:35 bubba [ 223.892621] #0: Jul 21 11:26:35 bubba (&buffer->mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] sysfs_write_file+0x37/0x13f Jul 21 11:26:35 bubba [ 223.893359] #1: Jul 21 11:26:35 bubba (s_active Jul 21 11:26:35 bubba ){++++.+} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] sysfs_write_file+0xe2/0x13f Jul 21 11:26:35 bubba [ 223.894094] #2: Jul 21 11:26:35 bubba (param_lock Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] param_attr_store+0x36/0x62 Jul 21 11:26:35 bubba [ 223.894835] #3: Jul 21 11:26:35 bubba (ft_mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] fcoe_transport_destroy+0x1e/0x110 [libfcoe] Jul 21 11:26:35 bubba [ 223.895574] #4: Jul 21 11:26:35 bubba (fcoe_config_mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] fcoe_destroy+0x18/0x72 [fcoe] Jul 21 11:26:35 bubba [ 223.896314] #5: Jul 21 11:26:35 bubba (rtnl_mutex Jul 21 11:26:35 bubba ){+.+.+.} Jul 21 11:26:35 bubba , at: Jul 21 11:26:35 bubba [] rtnl_lock+0x12/0x14 Jul 21 11:26:35 bubba [ 223.897047] Jul 21 11:26:35 bubba [ 223.897048] stack backtrace: Jul 21 11:26:35 bubba [ 223.897578] Pid: 3464, comm: lockdeptest.sh Not tainted 3.0.0-rc7+ #1 Jul 21 11:26:35 bubba [ 223.897853] Call Trace: Jul 21 11:26:35 bubba [ 223.898128] [] print_circular_bug+0x1f8/0x209 Jul 21 11:26:35 bubba [ 223.898416] [] __lock_acquire+0xb1d/0xe2c Jul 21 11:26:35 bubba [ 223.898699] [] ? wait_on_cpu_work+0xe6/0xe6 Jul 21 11:26:35 bubba [ 223.898982] [] lock_acquire+0xd2/0xf7 Jul 21 11:26:35 bubba [ 223.899263] [] ? wait_on_cpu_work+0xe6/0xe6 Jul 21 11:26:35 bubba [ 223.899547] [] ? mod_timer+0x8f/0x98 Jul 21 11:26:35 bubba [ 223.899827] [] wait_on_work+0x50/0xbd Jul 21 11:26:35 bubba [ 223.900108] [] ? wait_on_cpu_work+0xe6/0xe6 Jul 21 11:26:35 bubba [ 223.900390] [] __cancel_work_timer+0xb6/0xf4 Jul 21 11:26:35 bubba [ 223.900671] [] cancel_work_sync+0xb/0xd Jul 21 11:26:35 bubba [ 223.900953] [] fcoe_ctlr_destroy+0x1d/0x67 [libfcoe] Jul 21 11:26:35 bubba [ 223.901237] [] fcoe_interface_release+0x21/0x45 [fcoe] Jul 21 11:26:35 bubba [ 223.901522] [] ? fcoe_enable+0x6b/0x6b [fcoe] Jul 21 11:26:35 bubba [ 223.901803] [] kref_put+0x43/0x4d Jul 21 11:26:35 bubba [ 223.902083] [] fcoe_interface_put+0x17/0x19 [fcoe] Jul 21 11:26:35 bubba [ 223.902367] [] fcoe_interface_cleanup+0x188/0x193 [fcoe] Jul 21 11:26:35 bubba [ 223.902653] [] ? mutex_lock_nested+0x3b/0x40 Jul 21 11:26:35 bubba [ 223.902939] [] fcoe_destroy+0x52/0x72 [fcoe] Jul 21 11:26:35 bubba [ 223.903223] [] fcoe_transport_destroy+0xab/0x110 [libfcoe] Jul 21 11:26:35 bubba [ 223.903508] [] param_attr_store+0x43/0x62 Jul 21 11:26:35 bubba [ 223.903792] [] module_attr_store+0x21/0x25 Jul 21 11:26:35 bubba [ 223.904075] [] sysfs_write_file+0x103/0x13f Jul 21 11:26:35 bubba [ 223.904357] [] vfs_write+0xa7/0xfa Jul 21 11:26:35 bubba [ 223.904642] [] ? fget_light+0x35/0x96 Jul 21 11:26:35 bubba [ 223.904923] [] sys_write+0x45/0x69 Jul 21 11:26:35 bubba [ 223.905204] [] system_call_fastpath+0x16/0x1b Jul 21 11:26:36 bubba [ 223.964438] ixgbe 0000:05:00.0: eth3: detected SFP+: 5 Jul 21 11:26:37 bubba [ 225.196702] ixgbe 0000:05:00.0: eth3: NIC Link is Up 10 Gbps, Flow Control: None Signed-off-by: Robert Love Tested-by: Ross Brattain Reviewed-by: Yi Zou Signed-off-by: James Bottomley --- drivers/scsi/fcoe/fcoe.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c index ba710e350ac5..5d0e9a24ae94 100644 --- a/drivers/scsi/fcoe/fcoe.c +++ b/drivers/scsi/fcoe/fcoe.c @@ -432,6 +432,8 @@ void fcoe_interface_cleanup(struct fcoe_interface *fcoe) u8 flogi_maddr[ETH_ALEN]; const struct net_device_ops *ops; + rtnl_lock(); + /* * Don't listen for Ethernet packets anymore. * synchronize_net() ensures that the packet handlers are not running @@ -461,6 +463,8 @@ void fcoe_interface_cleanup(struct fcoe_interface *fcoe) " specific feature for LLD.\n"); } + rtnl_unlock(); + /* Release the self-reference taken during fcoe_interface_create() */ fcoe_interface_put(fcoe); } @@ -1951,11 +1955,8 @@ static void fcoe_destroy_work(struct work_struct *work) fcoe_if_destroy(port->lport); /* Do not tear down the fcoe interface for NPIV port */ - if (!npiv) { - rtnl_lock(); + if (!npiv) fcoe_interface_cleanup(fcoe); - rtnl_unlock(); - } mutex_unlock(&fcoe_config_mutex); } @@ -2009,8 +2010,9 @@ static int fcoe_create(struct net_device *netdev, enum fip_state fip_mode) printk(KERN_ERR "fcoe: Failed to create interface (%s)\n", netdev->name); rc = -EIO; + rtnl_unlock(); fcoe_interface_cleanup(fcoe); - goto out_nodev; + goto out_nortnl; } /* Make this the "master" N_Port */ @@ -2027,6 +2029,7 @@ static int fcoe_create(struct net_device *netdev, enum fip_state fip_mode) out_nodev: rtnl_unlock(); +out_nortnl: mutex_unlock(&fcoe_config_mutex); return rc; } From 77a2b73a7805a3c6a473b6741aa514ef40295d26 Mon Sep 17 00:00:00 2001 From: Vasu Dev Date: Thu, 25 Aug 2011 12:40:52 -0700 Subject: [PATCH 22/25] [SCSI] libfc: fix fc_eh_host_reset Current fc_eh_host_reset leaves lport offline permanently due to FLOGI response getting handled by LOGO response from last reset as both had same exchange id. So fix this by having end to end exches clean-up using exchange abort along exches reset done from fc_eh_host_reset. This would avoid exchanges collision between the sessions across the reset. In this case implicit login should have done that but no aborting support for FIP frames, so just wait till lport->r_a_tov before restarting next flogi to ensure all exchanges are good to use again for next session. Below is the trace of LOGO from older session coming ahead of FLOGI response with same exche id 0x203:- 617 86.435165 4e.00.0b -> ff.ff.fc FC ELS LOGO 0x203 618 86.435195 4e.00.0b -> b6.02.00 FC ELS LOGO 0x213 619 86.435220 4e.00.0b -> 18.03.00 FC ELS LOGO 0x223 620 86.435244 4e.00.0b -> 18.02.00 FC ELS LOGO 0x233 621 86.435267 4e.00.0b -> 18.01.00 FC ELS LOGO 0x243 622 86.435349 00.00.00 -> ff.ff.fe FC ELS FLOGI 0x203 623 86.435549 ff.ff.fc -> 4e.00.0b FC ELS ACC (LOGO) 0x203 624 86.438721 ff.ff.fe -> 4e.00.0b FC ELS ACC (FLOGI) 0x203 625 86.442059 18.03.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x223 626 86.443683 b6.02.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x213 627 86.447693 18.01.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x243 628 86.453499 18.02.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x233 Signed-off-by: Vasu Dev Tested-by: Ross Brattain Reviewed-by: Yi Zou Signed-off-by: Robert Love Signed-off-by: James Bottomley --- drivers/scsi/libfc/fc_exch.c | 51 ++++++++++++++++++++++++----------- drivers/scsi/libfc/fc_lport.c | 11 +++++++- 2 files changed, 45 insertions(+), 17 deletions(-) diff --git a/drivers/scsi/libfc/fc_exch.c b/drivers/scsi/libfc/fc_exch.c index 01ff082dc34c..744fefe81341 100644 --- a/drivers/scsi/libfc/fc_exch.c +++ b/drivers/scsi/libfc/fc_exch.c @@ -494,6 +494,9 @@ static int fc_seq_send(struct fc_lport *lport, struct fc_seq *sp, */ error = lport->tt.frame_send(lport, fp); + if (fh->fh_type == FC_TYPE_BLS) + return error; + /* * Update the exchange and sequence flags, * assuming all frames for the sequence have been sent. @@ -575,42 +578,35 @@ static void fc_seq_set_resp(struct fc_seq *sp, } /** - * fc_seq_exch_abort() - Abort an exchange and sequence - * @req_sp: The sequence to be aborted + * fc_exch_abort_locked() - Abort an exchange + * @ep: The exchange to be aborted * @timer_msec: The period of time to wait before aborting * - * Generally called because of a timeout or an abort from the upper layer. + * Locking notes: Called with exch lock held + * + * Return value: 0 on success else error code */ -static int fc_seq_exch_abort(const struct fc_seq *req_sp, - unsigned int timer_msec) +static int fc_exch_abort_locked(struct fc_exch *ep, + unsigned int timer_msec) { struct fc_seq *sp; - struct fc_exch *ep; struct fc_frame *fp; int error; - ep = fc_seq_exch(req_sp); - - spin_lock_bh(&ep->ex_lock); if (ep->esb_stat & (ESB_ST_COMPLETE | ESB_ST_ABNORMAL) || - ep->state & (FC_EX_DONE | FC_EX_RST_CLEANUP)) { - spin_unlock_bh(&ep->ex_lock); + ep->state & (FC_EX_DONE | FC_EX_RST_CLEANUP)) return -ENXIO; - } /* * Send the abort on a new sequence if possible. */ sp = fc_seq_start_next_locked(&ep->seq); - if (!sp) { - spin_unlock_bh(&ep->ex_lock); + if (!sp) return -ENOMEM; - } ep->esb_stat |= ESB_ST_SEQ_INIT | ESB_ST_ABNORMAL; if (timer_msec) fc_exch_timer_set_locked(ep, timer_msec); - spin_unlock_bh(&ep->ex_lock); /* * If not logged into the fabric, don't send ABTS but leave @@ -632,6 +628,28 @@ static int fc_seq_exch_abort(const struct fc_seq *req_sp, return error; } +/** + * fc_seq_exch_abort() - Abort an exchange and sequence + * @req_sp: The sequence to be aborted + * @timer_msec: The period of time to wait before aborting + * + * Generally called because of a timeout or an abort from the upper layer. + * + * Return value: 0 on success else error code + */ +static int fc_seq_exch_abort(const struct fc_seq *req_sp, + unsigned int timer_msec) +{ + struct fc_exch *ep; + int error; + + ep = fc_seq_exch(req_sp); + spin_lock_bh(&ep->ex_lock); + error = fc_exch_abort_locked(ep, timer_msec); + spin_unlock_bh(&ep->ex_lock); + return error; +} + /** * fc_exch_timeout() - Handle exchange timer expiration * @work: The work_struct identifying the exchange that timed out @@ -1715,6 +1733,7 @@ static void fc_exch_reset(struct fc_exch *ep) int rc = 1; spin_lock_bh(&ep->ex_lock); + fc_exch_abort_locked(ep, 0); ep->state |= FC_EX_RST_CLEANUP; if (cancel_delayed_work(&ep->timeout_work)) atomic_dec(&ep->ex_refcnt); /* drop hold for timer */ diff --git a/drivers/scsi/libfc/fc_lport.c b/drivers/scsi/libfc/fc_lport.c index e55ed9cf23fb..628f347404f9 100644 --- a/drivers/scsi/libfc/fc_lport.c +++ b/drivers/scsi/libfc/fc_lport.c @@ -88,6 +88,7 @@ */ #include +#include #include #include @@ -1029,8 +1030,16 @@ static void fc_lport_enter_reset(struct fc_lport *lport) FCH_EVT_LIPRESET, 0); fc_vports_linkchange(lport); fc_lport_reset_locked(lport); - if (lport->link_up) + if (lport->link_up) { + /* + * Wait upto resource allocation time out before + * doing re-login since incomplete FIP exchanged + * from last session may collide with exchanges + * in new session. + */ + msleep(lport->r_a_tov); fc_lport_enter_flogi(lport); + } } /** From 21cc0bd3a9e524b44a4f0ff05ac612aa0ff1a26e Mon Sep 17 00:00:00 2001 From: Vasu Dev Date: Thu, 25 Aug 2011 12:40:57 -0700 Subject: [PATCH 23/25] [SCSI] libfc: block SCSI eh thread for blocked rports Call fc_block_scsi_eh() in all fcoe eh to blocks the scsi_eh thread for blocked rports. Signed-off-by: Vasu Dev Tested-by: Ross Brattain Reviewed-by: Yi Zou Signed-off-by: Robert Love Signed-off-by: James Bottomley --- drivers/scsi/libfc/fc_fcp.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/libfc/fc_fcp.c b/drivers/scsi/libfc/fc_fcp.c index afb63c843144..4c41ee816f0b 100644 --- a/drivers/scsi/libfc/fc_fcp.c +++ b/drivers/scsi/libfc/fc_fcp.c @@ -2019,6 +2019,11 @@ int fc_eh_abort(struct scsi_cmnd *sc_cmd) struct fc_fcp_internal *si; int rc = FAILED; unsigned long flags; + int rval; + + rval = fc_block_scsi_eh(sc_cmd); + if (rval) + return rval; lport = shost_priv(sc_cmd->device->host); if (lport->state != LPORT_ST_READY) @@ -2068,9 +2073,9 @@ int fc_eh_device_reset(struct scsi_cmnd *sc_cmd) int rc = FAILED; int rval; - rval = fc_remote_port_chkready(rport); + rval = fc_block_scsi_eh(sc_cmd); if (rval) - goto out; + return rval; lport = shost_priv(sc_cmd->device->host); @@ -2116,6 +2121,8 @@ int fc_eh_host_reset(struct scsi_cmnd *sc_cmd) FC_SCSI_DBG(lport, "Resetting host\n"); + fc_block_scsi_eh(sc_cmd); + lport->tt.lport_reset(lport); wait_tmo = jiffies + FC_HOST_RESET_TIMEOUT; while (!fc_fcp_lport_queue_ready(lport) && time_before(jiffies, From 3ee17f59c5378af8d245f82498e3919b7de2ab40 Mon Sep 17 00:00:00 2001 From: Yi Zou Date: Thu, 25 Aug 2011 12:41:03 -0700 Subject: [PATCH 24/25] [SCSI] libfc: fix referencing to fc_fcp_pkt from the frame pointer via fr_fsp() In commit 6a716a8, while releasing the DDP context in case frame_send() failed, the frame may already be freed, so we should store the pointer to fc_fcp_pkt and release the DDP context using the locally stored fsp instead of getting fsp from the fr_fsp(fp) on a frame. Signed-off-by: Yi Zou Reported-by: Bhanu Prakash Gollapudi Tested-by: Ross Brattain Signed-off-by: Robert Love Signed-off-by: James Bottomley --- drivers/scsi/libfc/fc_exch.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/libfc/fc_exch.c b/drivers/scsi/libfc/fc_exch.c index 744fefe81341..d261e982a2fa 100644 --- a/drivers/scsi/libfc/fc_exch.c +++ b/drivers/scsi/libfc/fc_exch.c @@ -1981,6 +1981,7 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport, struct fc_exch *ep; struct fc_seq *sp = NULL; struct fc_frame_header *fh; + struct fc_fcp_pkt *fsp = NULL; int rc = 1; ep = fc_exch_alloc(lport, fp); @@ -2003,8 +2004,10 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport, fc_exch_setup_hdr(ep, fp, ep->f_ctl); sp->cnt++; - if (ep->xid <= lport->lro_xid && fh->fh_r_ctl == FC_RCTL_DD_UNSOL_CMD) + if (ep->xid <= lport->lro_xid && fh->fh_r_ctl == FC_RCTL_DD_UNSOL_CMD) { + fsp = fr_fsp(fp); fc_fcp_ddp_setup(fr_fsp(fp), ep->xid); + } if (unlikely(lport->tt.frame_send(lport, fp))) goto err; @@ -2018,7 +2021,8 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport, spin_unlock_bh(&ep->ex_lock); return sp; err: - fc_fcp_ddp_done(fr_fsp(fp)); + if (fsp) + fc_fcp_ddp_done(fsp); rc = fc_exch_done_locked(ep); spin_unlock_bh(&ep->ex_lock); if (!rc) From 610602f369b4c810c9df05e431abd38f38cb8e0d Mon Sep 17 00:00:00 2001 From: Eddie Wai Date: Fri, 26 Aug 2011 11:16:47 -0700 Subject: [PATCH 25/25] [SCSI] bnx2i: Fixed the endian on TTT for NOP out transmission The iscsi_nopout task's TTT is defined as __be32 while the DMA memory to the chip is CPU specific. This creates a problem for unsolicited NOP-In responses where the TTT is not the RESERVED tag of 0xFFs. This patch adds a call to be32_to_cpu for the TTT specified. Signed-off-by: Eddie Wai Signed-off-by: James Bottomley --- drivers/scsi/bnx2i/bnx2i_hwi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/bnx2i/bnx2i_hwi.c b/drivers/scsi/bnx2i/bnx2i_hwi.c index 9ae80cd5953b..dba72a4e6a1c 100644 --- a/drivers/scsi/bnx2i/bnx2i_hwi.c +++ b/drivers/scsi/bnx2i/bnx2i_hwi.c @@ -563,7 +563,7 @@ int bnx2i_send_iscsi_nopout(struct bnx2i_conn *bnx2i_conn, nopout_wqe->itt = ((u16)task->itt | (ISCSI_TASK_TYPE_MPATH << ISCSI_TMF_REQUEST_TYPE_SHIFT)); - nopout_wqe->ttt = nopout_hdr->ttt; + nopout_wqe->ttt = be32_to_cpu(nopout_hdr->ttt); nopout_wqe->flags = 0; if (!unsol) nopout_wqe->flags = ISCSI_NOP_OUT_REQUEST_LOCAL_COMPLETION;