diff options
author | Peter Oberparleiter <oberpar@linux.ibm.com> | 2021-08-13 15:05:05 +0200 |
---|---|---|
committer | Heiko Carstens <hca@linux.ibm.com> | 2021-08-25 11:03:35 +0200 |
commit | 70aa5d39826528e77f5595a5f9297d919112d396 (patch) | |
tree | 42af2f47b3bb474a6a8213a69d8ba268babe70c7 | |
parent | s390/debug: add early tracing support (diff) | |
download | linux-70aa5d39826528e77f5595a5f9297d919112d396.tar.xz linux-70aa5d39826528e77f5595a5f9297d919112d396.zip |
s390/sclp: add tracing of SCLP interactions
Add tracing of interactions between the SCLP base driver, firmware and
other drivers to support problem determination in case of SCLP-related
issues.
For that purpose this patch introduces two new s390dbf debug areas:
- sclp: An abbreviated log of all common interactions
- sclp_err: A full log of failed or abnormal interactions
Tracing of full SCCB contents can be enabled for the sclp area by
setting its debug level to maximum (6).
Overview of added trace events:
* Firmware interaction:
- SRV1: Service call about to be issued
- SRV2: Service call was issued
- INT: Interrupt received
* Driver interaction:
- RQAD: Request was added
- RQOK: Request success
- RQAB: Request aborted
- RQTM: Request timed out
- REG: Event listener registered
- UREG: Event listener unregistered
- EVNT: Event callback
- STCG: State-change callback
* Abnormal events:
- TMO: A timeout occurred
- UNEX: Unexpected SCCB completion
* Other (not traced at default level):
- SYN1: Synchronous wait start
- SYN2: Synchronous wait end
Since the SCLP interface is used by console drivers this patch also
moves s390dbf printks outside the critical section protected by debug
area locks to prevent a potential deadlock that would otherwise be
introduced between console_owner --> sclp_lock --> sclp_debug.lock.
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
-rw-r--r-- | arch/s390/kernel/debug.c | 16 | ||||
-rw-r--r-- | drivers/s390/char/sclp.c | 230 |
2 files changed, 236 insertions, 10 deletions
diff --git a/arch/s390/kernel/debug.c b/arch/s390/kernel/debug.c index da4d67736daa..4331c7e6e1c0 100644 --- a/arch/s390/kernel/debug.c +++ b/arch/s390/kernel/debug.c @@ -833,16 +833,17 @@ void debug_set_level(debug_info_t *id, int new_level) if (!id) return; - spin_lock_irqsave(&id->lock, flags); + if (new_level == DEBUG_OFF_LEVEL) { - id->level = DEBUG_OFF_LEVEL; pr_info("%s: switched off\n", id->name); } else if ((new_level > DEBUG_MAX_LEVEL) || (new_level < 0)) { pr_info("%s: level %i is out of range (%i - %i)\n", id->name, new_level, 0, DEBUG_MAX_LEVEL); - } else { - id->level = new_level; + return; } + + spin_lock_irqsave(&id->lock, flags); + id->level = new_level; spin_unlock_irqrestore(&id->lock, flags); } EXPORT_SYMBOL(debug_set_level); @@ -1208,16 +1209,17 @@ int debug_register_view(debug_info_t *id, struct debug_view *view) break; } if (i == DEBUG_MAX_VIEWS) { - pr_err("Registering view %s/%s would exceed the maximum " - "number of views %i\n", id->name, view->name, i); rc = -1; } else { id->views[i] = view; id->debugfs_entries[i] = pde; } spin_unlock_irqrestore(&id->lock, flags); - if (rc) + if (rc) { + pr_err("Registering view %s/%s would exceed the maximum " + "number of views %i\n", id->name, view->name, i); debugfs_remove(pde); + } out: return rc; } diff --git a/drivers/s390/char/sclp.c b/drivers/s390/char/sclp.c index 792b4bfa6d9a..b4b84e3e0949 100644 --- a/drivers/s390/char/sclp.c +++ b/drivers/s390/char/sclp.c @@ -21,11 +21,30 @@ #include <linux/platform_device.h> #include <asm/types.h> #include <asm/irq.h> +#include <asm/debug.h> #include "sclp.h" #define SCLP_HEADER "sclp: " +struct sclp_trace_entry { + char id[4]; + u32 a; + u64 b; +}; + +#define SCLP_TRACE_ENTRY_SIZE sizeof(struct sclp_trace_entry) +#define SCLP_TRACE_MAX_SIZE 128 +#define SCLP_TRACE_EVENT_MAX_SIZE 64 + +/* Debug trace area intended for all entries in abbreviated form. */ +DEFINE_STATIC_DEBUG_INFO(sclp_debug, "sclp", 8, 1, SCLP_TRACE_ENTRY_SIZE, + &debug_hex_ascii_view); + +/* Error trace area intended for full entries relating to failed requests. */ +DEFINE_STATIC_DEBUG_INFO(sclp_debug_err, "sclp_err", 4, 1, + SCLP_TRACE_ENTRY_SIZE, &debug_hex_ascii_view); + /* Lock to protect internal data consistency. */ static DEFINE_SPINLOCK(sclp_lock); @@ -54,6 +73,114 @@ int sclp_console_drop = 1; /* Number of times the console dropped buffer pages */ unsigned long sclp_console_full; +/* The currently active SCLP command word. */ +static sclp_cmdw_t active_cmd; + +static inline void sclp_trace(int prio, char *id, u32 a, u64 b, bool err) +{ + struct sclp_trace_entry e; + + memset(&e, 0, sizeof(e)); + strncpy(e.id, id, sizeof(e.id)); + e.a = a; + e.b = b; + debug_event(&sclp_debug, prio, &e, sizeof(e)); + if (err) + debug_event(&sclp_debug_err, 0, &e, sizeof(e)); +} + +static inline int no_zeroes_len(void *data, int len) +{ + char *d = data; + + /* Minimize trace area usage by not tracing trailing zeroes. */ + while (len > SCLP_TRACE_ENTRY_SIZE && d[len - 1] == 0) + len--; + + return len; +} + +static inline void sclp_trace_bin(int prio, void *d, int len, int errlen) +{ + debug_event(&sclp_debug, prio, d, no_zeroes_len(d, len)); + if (errlen) + debug_event(&sclp_debug_err, 0, d, no_zeroes_len(d, errlen)); +} + +static inline int abbrev_len(sclp_cmdw_t cmd, struct sccb_header *sccb) +{ + struct evbuf_header *evbuf = (struct evbuf_header *)(sccb + 1); + int len = sccb->length, limit = SCLP_TRACE_MAX_SIZE; + + /* Full SCCB tracing if debug level is set to max. */ + if (sclp_debug.level == DEBUG_MAX_LEVEL) + return len; + + /* Minimal tracing for console writes. */ + if (cmd == SCLP_CMDW_WRITE_EVENT_DATA && + (evbuf->type == EVTYP_MSG || evbuf->type == EVTYP_VT220MSG)) + limit = SCLP_TRACE_ENTRY_SIZE; + + return min(len, limit); +} + +static inline void sclp_trace_sccb(int prio, char *id, u32 a, u64 b, + sclp_cmdw_t cmd, struct sccb_header *sccb, + bool err) +{ + sclp_trace(prio, id, a, b, err); + if (sccb) { + sclp_trace_bin(prio + 1, sccb, abbrev_len(cmd, sccb), + err ? sccb->length : 0); + } +} + +static inline void sclp_trace_evbuf(int prio, char *id, u32 a, u64 b, + struct evbuf_header *evbuf, bool err) +{ + sclp_trace(prio, id, a, b, err); + sclp_trace_bin(prio + 1, evbuf, + min((int)evbuf->length, (int)SCLP_TRACE_EVENT_MAX_SIZE), + err ? evbuf->length : 0); +} + +static inline void sclp_trace_req(int prio, char *id, struct sclp_req *req, + bool err) +{ + struct sccb_header *sccb = req->sccb; + union { + struct { + u16 status; + u16 response; + u16 timeout; + u16 start_count; + }; + u64 b; + } summary; + + summary.status = req->status; + summary.response = sccb ? sccb->response_code : 0; + summary.timeout = (u16)req->queue_timeout; + summary.start_count = (u16)req->start_count; + + sclp_trace(prio, id, (u32)(addr_t)sccb, summary.b, err); +} + +static inline void sclp_trace_register(int prio, char *id, u32 a, u64 b, + struct sclp_register *reg) +{ + struct { + u64 receive; + u64 send; + } d; + + d.receive = reg->receive_mask; + d.send = reg->send_mask; + + sclp_trace(prio, id, a, b, false); + sclp_trace_bin(prio, &d, sizeof(d), 0); +} + static int __init sclp_setup_console_pages(char *str) { int pages, rc; @@ -162,6 +289,9 @@ static void sclp_request_timeout(bool force_restart) { unsigned long flags; + /* TMO: A timeout occurred (a=force_restart) */ + sclp_trace(2, "TMO", force_restart, 0, true); + spin_lock_irqsave(&sclp_lock, flags); if (force_restart) { if (sclp_running_state == sclp_running_state_running) { @@ -237,6 +367,12 @@ static void sclp_req_queue_timeout(struct timer_list *unused) do { req = __sclp_req_queue_remove_expired_req(); + + if (req) { + /* RQTM: Request timed out (a=sccb, b=summary) */ + sclp_trace_req(2, "RQTM", req, true); + } + if (req && req->callback) req->callback(req, req->callback_data); } while (req); @@ -248,6 +384,25 @@ static void sclp_req_queue_timeout(struct timer_list *unused) spin_unlock_irqrestore(&sclp_lock, flags); } +static int sclp_service_call_trace(sclp_cmdw_t command, void *sccb) +{ + static u64 srvc_count; + int rc; + + /* SRV1: Service call about to be issued (a=command, b=sccb address) */ + sclp_trace_sccb(0, "SRV1", command, (u64)sccb, command, sccb, false); + + rc = sclp_service_call(command, sccb); + + /* SRV2: Service call was issued (a=rc, b=SRVC sequence number) */ + sclp_trace(0, "SRV2", -rc, ++srvc_count, rc != 0); + + if (rc == 0) + active_cmd = command; + + return rc; +} + /* Try to start a request. Return zero if the request was successfully * started or if it will be started at a later time. Return non-zero otherwise. * Called while sclp_lock is locked. */ @@ -259,7 +414,7 @@ __sclp_start_request(struct sclp_req *req) if (sclp_running_state != sclp_running_state_idle) return 0; del_timer(&sclp_request_timer); - rc = sclp_service_call(req->command, req->sccb); + rc = sclp_service_call_trace(req->command, req->sccb); req->start_count++; if (rc == 0) { @@ -309,6 +464,10 @@ sclp_process_queue(void) } /* Post-processing for aborted request */ list_del(&req->list); + + /* RQAB: Request aborted (a=sccb, b=summary) */ + sclp_trace_req(2, "RQAB", req, true); + if (req->callback) { spin_unlock_irqrestore(&sclp_lock, flags); req->callback(req, req->callback_data); @@ -341,6 +500,10 @@ sclp_add_request(struct sclp_req *req) spin_unlock_irqrestore(&sclp_lock, flags); return -EIO; } + + /* RQAD: Request was added (a=sccb, b=caller) */ + sclp_trace(2, "RQAD", (u32)(addr_t)req->sccb, _RET_IP_, false); + req->status = SCLP_REQ_QUEUED; req->start_count = 0; list_add_tail(&req->list, &sclp_req_queue); @@ -394,6 +557,11 @@ sclp_dispatch_evbufs(struct sccb_header *sccb) else reg = NULL; } + + /* EVNT: Event callback (b=receiver) */ + sclp_trace_evbuf(2, "EVNT", 0, reg ? (u64)reg->receiver_fn : 0, + evbuf, !reg); + if (reg && reg->receiver_fn) { spin_unlock_irqrestore(&sclp_lock, flags); reg->receiver_fn(evbuf); @@ -455,6 +623,30 @@ __sclp_find_req(u32 sccb) return NULL; } +static bool ok_response(u32 sccb_int, sclp_cmdw_t cmd) +{ + struct sccb_header *sccb = (struct sccb_header *)(addr_t)sccb_int; + struct evbuf_header *evbuf; + u16 response; + + if (!sccb) + return true; + + /* Check SCCB response. */ + response = sccb->response_code & 0xff; + if (response != 0x10 && response != 0x20) + return false; + + /* Check event-processed flag on outgoing events. */ + if (cmd == SCLP_CMDW_WRITE_EVENT_DATA) { + evbuf = (struct evbuf_header *)(sccb + 1); + if (!(evbuf->flags & 0x80)) + return false; + } + + return true; +} + /* Handler for external interruption. Perform request post-processing. * Prepare read event data request if necessary. Start processing of next * request on queue. */ @@ -469,6 +661,12 @@ static void sclp_interrupt_handler(struct ext_code ext_code, spin_lock(&sclp_lock); finished_sccb = param32 & 0xfffffff8; evbuf_pending = param32 & 0x3; + + /* INT: Interrupt received (a=intparm, b=cmd) */ + sclp_trace_sccb(0, "INT", param32, active_cmd, active_cmd, + (struct sccb_header *)(addr_t)finished_sccb, + !ok_response(finished_sccb, active_cmd)); + if (finished_sccb) { del_timer(&sclp_request_timer); sclp_running_state = sclp_running_state_reset_pending; @@ -477,13 +675,21 @@ static void sclp_interrupt_handler(struct ext_code ext_code, /* Request post-processing */ list_del(&req->list); req->status = SCLP_REQ_DONE; + + /* RQOK: Request success (a=sccb, b=summary) */ + sclp_trace_req(2, "RQOK", req, false); + if (req->callback) { spin_unlock(&sclp_lock); req->callback(req, req->callback_data); spin_lock(&sclp_lock); } + } else { + /* UNEX: Unexpected SCCB completion (a=sccb address) */ + sclp_trace(0, "UNEX", finished_sccb, 0, true); } sclp_running_state = sclp_running_state_idle; + active_cmd = 0; } if (evbuf_pending && sclp_activation_state == sclp_activation_state_active) @@ -507,9 +713,13 @@ sclp_sync_wait(void) unsigned long long old_tick; unsigned long flags; unsigned long cr0, cr0_sync; + static u64 sync_count; u64 timeout; int irq_context; + /* SYN1: Synchronous wait start (a=runstate, b=sync count) */ + sclp_trace(4, "SYN1", sclp_running_state, ++sync_count, false); + /* We'll be disabling timer interrupts, so we need a custom timeout * mechanism */ timeout = 0; @@ -547,6 +757,9 @@ sclp_sync_wait(void) _local_bh_enable(); local_tick_enable(old_tick); local_irq_restore(flags); + + /* SYN2: Synchronous wait end (a=runstate, b=sync_count) */ + sclp_trace(4, "SYN2", sclp_running_state, sync_count, false); } EXPORT_SYMBOL(sclp_sync_wait); @@ -576,8 +789,13 @@ sclp_dispatch_state_change(void) reg = NULL; } spin_unlock_irqrestore(&sclp_lock, flags); - if (reg && reg->state_change_fn) + if (reg && reg->state_change_fn) { + /* STCG: State-change callback (b=callback) */ + sclp_trace(2, "STCG", 0, (u64)reg->state_change_fn, + false); + reg->state_change_fn(reg); + } } while (reg); } @@ -651,6 +869,9 @@ sclp_register(struct sclp_register *reg) sccb_mask_t send_mask; int rc; + /* REG: Event listener registered (b=caller) */ + sclp_trace_register(2, "REG", 0, _RET_IP_, reg); + rc = sclp_init(); if (rc) return rc; @@ -683,6 +904,9 @@ sclp_unregister(struct sclp_register *reg) { unsigned long flags; + /* UREG: Event listener unregistered (b=caller) */ + sclp_trace_register(2, "UREG", 0, _RET_IP_, reg); + spin_lock_irqsave(&sclp_lock, flags); list_del(®->list); spin_unlock_irqrestore(&sclp_lock, flags); @@ -932,7 +1156,7 @@ sclp_check_interface(void) for (retry = 0; retry <= SCLP_INIT_RETRY; retry++) { __sclp_make_init_req(0, 0); sccb = (struct init_sccb *) sclp_init_req.sccb; - rc = sclp_service_call(sclp_init_req.command, sccb); + rc = sclp_service_call_trace(sclp_init_req.command, sccb); if (rc == -EIO) break; sclp_init_req.status = SCLP_REQ_RUNNING; |