From cd479eccd2e057116d504852814402a1e68ead80 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Mon, 4 Mar 2019 12:33:28 +0100 Subject: [PATCH 1/6] s390: limit brk randomization to 32MB For a 64-bit process the randomization of the program break is quite large with 1GB. That is as big as the randomization of the anonymous mapping base, for a test case started with '/lib/ld64.so.1 ' it can happen that the heap is placed after the stack. To avoid this limit the program break randomization to 32MB for 64-bit and keep 8MB for 31-bit. Reported-by: Stefan Liebler Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/elf.h | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/arch/s390/include/asm/elf.h b/arch/s390/include/asm/elf.h index 7d22a474a040..f74639a05f0f 100644 --- a/arch/s390/include/asm/elf.h +++ b/arch/s390/include/asm/elf.h @@ -252,11 +252,14 @@ do { \ /* * Cache aliasing on the latest machines calls for a mapping granularity - * of 512KB. For 64-bit processes use a 512KB alignment and a randomization - * of up to 1GB. For 31-bit processes the virtual address space is limited, - * use no alignment and limit the randomization to 8MB. + * of 512KB for the anonymous mapping base. For 64-bit processes use a + * 512KB alignment and a randomization of up to 1GB. For 31-bit processes + * the virtual address space is limited, use no alignment and limit the + * randomization to 8MB. + * For the additional randomization of the program break use 32MB for + * 64-bit and 8MB for 31-bit. */ -#define BRK_RND_MASK (is_compat_task() ? 0x7ffUL : 0x3ffffUL) +#define BRK_RND_MASK (is_compat_task() ? 0x7ffUL : 0x1fffUL) #define MMAP_RND_MASK (is_compat_task() ? 0x7ffUL : 0x3ff80UL) #define MMAP_ALIGN_MASK (is_compat_task() ? 0 : 0x7fUL) #define STACK_RND_MASK MMAP_RND_MASK From 01396a374c3d31bc5f8b693026cfa9a657319624 Mon Sep 17 00:00:00 2001 From: Harald Freudenberger Date: Fri, 22 Feb 2019 17:24:11 +0100 Subject: [PATCH 2/6] s390/zcrypt: revisit ap device remove procedure Working with the vfio-ap driver let to some revisit of the way how an ap (queue) device is removed from the driver. With the current implementation all the cleanup was done before the driver even got notified about the removal. Now the ap queue removal is done in 3 steps: 1) A preparation step, all ap messages within the queue are flushed and so the driver does 'receive' them. Also a new state AP_STATE_REMOVE assigned to the queue makes sure there are no new messages queued in. 2) Now the driver's remove function is invoked and the driver should do the job of cleaning up it's internal administration lists or whatever. After 2) is done it is guaranteed, that the driver is not invoked any more. On the other hand the driver has to make sure that the APQN is not accessed any more after step 2 is complete. 3) Now the ap bus code does the job of total cleanup of the APQN. A reset with zero is triggered and the state of the queue goes to AP_STATE_UNBOUND. After step 3) is complete, the ap queue has no pending messages and the APQN is cleared and so there are no requests and replies lingering around in the firmware queue for this APQN. Also the interrupts are disabled. After these remove steps the ap queue device may be assigned to another driver. Stress testing this remove/probe procedure showed a problem with the correct module reference counting. The actual receive of an reply in the driver is done asynchronous with completions. So with a driver change on an ap queue the message flush triggers completions but the threads waiting for the completions may run at a time where the queue already has the new driver assigned. So the module_put() at receive time needs to be done on the driver module which queued the ap message. This change is also part of this patch. Signed-off-by: Harald Freudenberger Reviewed-by: Ingo Franzki Signed-off-by: Martin Schwidefsky --- drivers/s390/crypto/ap_bus.c | 9 ++++++++- drivers/s390/crypto/ap_bus.h | 2 ++ drivers/s390/crypto/ap_queue.c | 28 ++++++++++++++++++++++------ drivers/s390/crypto/zcrypt_api.c | 30 ++++++++++++++++++------------ 4 files changed, 50 insertions(+), 19 deletions(-) diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index e15816ff1265..033a1acabf48 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -810,11 +810,18 @@ static int ap_device_remove(struct device *dev) struct ap_device *ap_dev = to_ap_dev(dev); struct ap_driver *ap_drv = ap_dev->drv; + /* prepare ap queue device removal */ if (is_queue_dev(dev)) - ap_queue_remove(to_ap_queue(dev)); + ap_queue_prepare_remove(to_ap_queue(dev)); + + /* driver's chance to clean up gracefully */ if (ap_drv->remove) ap_drv->remove(ap_dev); + /* now do the ap queue device remove */ + if (is_queue_dev(dev)) + ap_queue_remove(to_ap_queue(dev)); + /* Remove queue/card from list of active queues/cards */ spin_lock_bh(&ap_list_lock); if (is_card_dev(dev)) diff --git a/drivers/s390/crypto/ap_bus.h b/drivers/s390/crypto/ap_bus.h index d0059eae5d94..15a98a673c5c 100644 --- a/drivers/s390/crypto/ap_bus.h +++ b/drivers/s390/crypto/ap_bus.h @@ -91,6 +91,7 @@ enum ap_state { AP_STATE_WORKING, AP_STATE_QUEUE_FULL, AP_STATE_SUSPEND_WAIT, + AP_STATE_REMOVE, /* about to be removed from driver */ AP_STATE_UNBOUND, /* momentary not bound to a driver */ AP_STATE_BORKED, /* broken */ NR_AP_STATES @@ -252,6 +253,7 @@ void ap_bus_force_rescan(void); void ap_queue_init_reply(struct ap_queue *aq, struct ap_message *ap_msg); struct ap_queue *ap_queue_create(ap_qid_t qid, int device_type); +void ap_queue_prepare_remove(struct ap_queue *aq); void ap_queue_remove(struct ap_queue *aq); void ap_queue_suspend(struct ap_device *ap_dev); void ap_queue_resume(struct ap_device *ap_dev); diff --git a/drivers/s390/crypto/ap_queue.c b/drivers/s390/crypto/ap_queue.c index ba261210c6da..6a340f2c3556 100644 --- a/drivers/s390/crypto/ap_queue.c +++ b/drivers/s390/crypto/ap_queue.c @@ -420,6 +420,10 @@ static ap_func_t *ap_jumptable[NR_AP_STATES][NR_AP_EVENTS] = { [AP_EVENT_POLL] = ap_sm_suspend_read, [AP_EVENT_TIMEOUT] = ap_sm_nop, }, + [AP_STATE_REMOVE] = { + [AP_EVENT_POLL] = ap_sm_nop, + [AP_EVENT_TIMEOUT] = ap_sm_nop, + }, [AP_STATE_UNBOUND] = { [AP_EVENT_POLL] = ap_sm_nop, [AP_EVENT_TIMEOUT] = ap_sm_nop, @@ -740,18 +744,31 @@ void ap_flush_queue(struct ap_queue *aq) } EXPORT_SYMBOL(ap_flush_queue); +void ap_queue_prepare_remove(struct ap_queue *aq) +{ + spin_lock_bh(&aq->lock); + /* flush queue */ + __ap_flush_queue(aq); + /* set REMOVE state to prevent new messages are queued in */ + aq->state = AP_STATE_REMOVE; + del_timer_sync(&aq->timeout); + spin_unlock_bh(&aq->lock); +} + void ap_queue_remove(struct ap_queue *aq) { - ap_flush_queue(aq); - del_timer_sync(&aq->timeout); - - /* reset with zero, also clears irq registration */ + /* + * all messages have been flushed and the state is + * AP_STATE_REMOVE. Now reset with zero which also + * clears the irq registration and move the state + * to AP_STATE_UNBOUND to signal that this queue + * is not used by any driver currently. + */ spin_lock_bh(&aq->lock); ap_zapq(aq->qid); aq->state = AP_STATE_UNBOUND; spin_unlock_bh(&aq->lock); } -EXPORT_SYMBOL(ap_queue_remove); void ap_queue_reinit_state(struct ap_queue *aq) { @@ -760,4 +777,3 @@ void ap_queue_reinit_state(struct ap_queue *aq) ap_wait(ap_sm_event(aq, AP_EVENT_POLL)); spin_unlock_bh(&aq->lock); } -EXPORT_SYMBOL(ap_queue_reinit_state); diff --git a/drivers/s390/crypto/zcrypt_api.c b/drivers/s390/crypto/zcrypt_api.c index eb93c2d27d0a..689c2af7026a 100644 --- a/drivers/s390/crypto/zcrypt_api.c +++ b/drivers/s390/crypto/zcrypt_api.c @@ -586,6 +586,7 @@ static inline bool zcrypt_check_queue(struct ap_perms *perms, int queue) static inline struct zcrypt_queue *zcrypt_pick_queue(struct zcrypt_card *zc, struct zcrypt_queue *zq, + struct module **pmod, unsigned int weight) { if (!zq || !try_module_get(zq->queue->ap_dev.drv->driver.owner)) @@ -595,15 +596,15 @@ static inline struct zcrypt_queue *zcrypt_pick_queue(struct zcrypt_card *zc, atomic_add(weight, &zc->load); atomic_add(weight, &zq->load); zq->request_count++; + *pmod = zq->queue->ap_dev.drv->driver.owner; return zq; } static inline void zcrypt_drop_queue(struct zcrypt_card *zc, struct zcrypt_queue *zq, + struct module *mod, unsigned int weight) { - struct module *mod = zq->queue->ap_dev.drv->driver.owner; - zq->request_count--; atomic_sub(weight, &zc->load); atomic_sub(weight, &zq->load); @@ -653,6 +654,7 @@ static long zcrypt_rsa_modexpo(struct ap_perms *perms, unsigned int weight, pref_weight; unsigned int func_code; int qid = 0, rc = -ENODEV; + struct module *mod; trace_s390_zcrypt_req(mex, TP_ICARSAMODEXPO); @@ -706,7 +708,7 @@ static long zcrypt_rsa_modexpo(struct ap_perms *perms, pref_weight = weight; } } - pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, weight); + pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, &mod, weight); spin_unlock(&zcrypt_list_lock); if (!pref_zq) { @@ -718,7 +720,7 @@ static long zcrypt_rsa_modexpo(struct ap_perms *perms, rc = pref_zq->ops->rsa_modexpo(pref_zq, mex); spin_lock(&zcrypt_list_lock); - zcrypt_drop_queue(pref_zc, pref_zq, weight); + zcrypt_drop_queue(pref_zc, pref_zq, mod, weight); spin_unlock(&zcrypt_list_lock); out: @@ -735,6 +737,7 @@ static long zcrypt_rsa_crt(struct ap_perms *perms, unsigned int weight, pref_weight; unsigned int func_code; int qid = 0, rc = -ENODEV; + struct module *mod; trace_s390_zcrypt_req(crt, TP_ICARSACRT); @@ -788,7 +791,7 @@ static long zcrypt_rsa_crt(struct ap_perms *perms, pref_weight = weight; } } - pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, weight); + pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, &mod, weight); spin_unlock(&zcrypt_list_lock); if (!pref_zq) { @@ -800,7 +803,7 @@ static long zcrypt_rsa_crt(struct ap_perms *perms, rc = pref_zq->ops->rsa_modexpo_crt(pref_zq, crt); spin_lock(&zcrypt_list_lock); - zcrypt_drop_queue(pref_zc, pref_zq, weight); + zcrypt_drop_queue(pref_zc, pref_zq, mod, weight); spin_unlock(&zcrypt_list_lock); out: @@ -819,6 +822,7 @@ static long _zcrypt_send_cprb(struct ap_perms *perms, unsigned int func_code; unsigned short *domain; int qid = 0, rc = -ENODEV; + struct module *mod; trace_s390_zcrypt_req(xcRB, TB_ZSECSENDCPRB); @@ -865,7 +869,7 @@ static long _zcrypt_send_cprb(struct ap_perms *perms, pref_weight = weight; } } - pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, weight); + pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, &mod, weight); spin_unlock(&zcrypt_list_lock); if (!pref_zq) { @@ -881,7 +885,7 @@ static long _zcrypt_send_cprb(struct ap_perms *perms, rc = pref_zq->ops->send_cprb(pref_zq, xcRB, &ap_msg); spin_lock(&zcrypt_list_lock); - zcrypt_drop_queue(pref_zc, pref_zq, weight); + zcrypt_drop_queue(pref_zc, pref_zq, mod, weight); spin_unlock(&zcrypt_list_lock); out: @@ -932,6 +936,7 @@ static long zcrypt_send_ep11_cprb(struct ap_perms *perms, unsigned int func_code; struct ap_message ap_msg; int qid = 0, rc = -ENODEV; + struct module *mod; trace_s390_zcrypt_req(xcrb, TP_ZSENDEP11CPRB); @@ -1000,7 +1005,7 @@ static long zcrypt_send_ep11_cprb(struct ap_perms *perms, pref_weight = weight; } } - pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, weight); + pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, &mod, weight); spin_unlock(&zcrypt_list_lock); if (!pref_zq) { @@ -1012,7 +1017,7 @@ static long zcrypt_send_ep11_cprb(struct ap_perms *perms, rc = pref_zq->ops->send_ep11_cprb(pref_zq, xcrb, &ap_msg); spin_lock(&zcrypt_list_lock); - zcrypt_drop_queue(pref_zc, pref_zq, weight); + zcrypt_drop_queue(pref_zc, pref_zq, mod, weight); spin_unlock(&zcrypt_list_lock); out_free: @@ -1033,6 +1038,7 @@ static long zcrypt_rng(char *buffer) struct ap_message ap_msg; unsigned int domain; int qid = 0, rc = -ENODEV; + struct module *mod; trace_s390_zcrypt_req(buffer, TP_HWRNGCPRB); @@ -1064,7 +1070,7 @@ static long zcrypt_rng(char *buffer) pref_weight = weight; } } - pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, weight); + pref_zq = zcrypt_pick_queue(pref_zc, pref_zq, &mod, weight); spin_unlock(&zcrypt_list_lock); if (!pref_zq) { @@ -1076,7 +1082,7 @@ static long zcrypt_rng(char *buffer) rc = pref_zq->ops->rng(pref_zq, buffer, &ap_msg); spin_lock(&zcrypt_list_lock); - zcrypt_drop_queue(pref_zc, pref_zq, weight); + zcrypt_drop_queue(pref_zc, pref_zq, mod, weight); spin_unlock(&zcrypt_list_lock); out: From 152e9b8676c6e788c6bff095c1eaae7b86df5003 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Wed, 6 Mar 2019 13:31:21 +0200 Subject: [PATCH 3/6] s390/vtime: steal time exponential moving average To be able to judge the current overcommitment ratio for a CPU add a lowcore field with the exponential moving average of the steal time. The average is updated every tick. Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/lowcore.h | 61 +++++++++++++++++---------------- arch/s390/kernel/smp.c | 3 +- arch/s390/kernel/vtime.c | 19 ++++++---- 3 files changed, 45 insertions(+), 38 deletions(-) diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h index cc0947e08b6f..5b9f10b1e55d 100644 --- a/arch/s390/include/asm/lowcore.h +++ b/arch/s390/include/asm/lowcore.h @@ -91,52 +91,53 @@ struct lowcore { __u64 hardirq_timer; /* 0x02e8 */ __u64 softirq_timer; /* 0x02f0 */ __u64 steal_timer; /* 0x02f8 */ - __u64 last_update_timer; /* 0x0300 */ - __u64 last_update_clock; /* 0x0308 */ - __u64 int_clock; /* 0x0310 */ - __u64 mcck_clock; /* 0x0318 */ - __u64 clock_comparator; /* 0x0320 */ - __u64 boot_clock[2]; /* 0x0328 */ + __u64 avg_steal_timer; /* 0x0300 */ + __u64 last_update_timer; /* 0x0308 */ + __u64 last_update_clock; /* 0x0310 */ + __u64 int_clock; /* 0x0318*/ + __u64 mcck_clock; /* 0x0320 */ + __u64 clock_comparator; /* 0x0328 */ + __u64 boot_clock[2]; /* 0x0330 */ /* Current process. */ - __u64 current_task; /* 0x0338 */ - __u64 kernel_stack; /* 0x0340 */ + __u64 current_task; /* 0x0340 */ + __u64 kernel_stack; /* 0x0348 */ /* Interrupt, DAT-off and restartstack. */ - __u64 async_stack; /* 0x0348 */ - __u64 nodat_stack; /* 0x0350 */ - __u64 restart_stack; /* 0x0358 */ + __u64 async_stack; /* 0x0350 */ + __u64 nodat_stack; /* 0x0358 */ + __u64 restart_stack; /* 0x0360 */ /* Restart function and parameter. */ - __u64 restart_fn; /* 0x0360 */ - __u64 restart_data; /* 0x0368 */ - __u64 restart_source; /* 0x0370 */ + __u64 restart_fn; /* 0x0368 */ + __u64 restart_data; /* 0x0370 */ + __u64 restart_source; /* 0x0378 */ /* Address space pointer. */ - __u64 kernel_asce; /* 0x0378 */ - __u64 user_asce; /* 0x0380 */ - __u64 vdso_asce; /* 0x0388 */ + __u64 kernel_asce; /* 0x0380 */ + __u64 user_asce; /* 0x0388 */ + __u64 vdso_asce; /* 0x0390 */ /* * The lpp and current_pid fields form a * 64-bit value that is set as program * parameter with the LPP instruction. */ - __u32 lpp; /* 0x0390 */ - __u32 current_pid; /* 0x0394 */ + __u32 lpp; /* 0x0398 */ + __u32 current_pid; /* 0x039c */ /* SMP info area */ - __u32 cpu_nr; /* 0x0398 */ - __u32 softirq_pending; /* 0x039c */ - __u32 preempt_count; /* 0x03a0 */ - __u32 spinlock_lockval; /* 0x03a4 */ - __u32 spinlock_index; /* 0x03a8 */ - __u32 fpu_flags; /* 0x03ac */ - __u64 percpu_offset; /* 0x03b0 */ - __u64 vdso_per_cpu_data; /* 0x03b8 */ - __u64 machine_flags; /* 0x03c0 */ - __u64 gmap; /* 0x03c8 */ - __u8 pad_0x03d0[0x0400-0x03d0]; /* 0x03d0 */ + __u32 cpu_nr; /* 0x03a0 */ + __u32 softirq_pending; /* 0x03a4 */ + __u32 preempt_count; /* 0x03a8 */ + __u32 spinlock_lockval; /* 0x03ac */ + __u32 spinlock_index; /* 0x03b0 */ + __u32 fpu_flags; /* 0x03b4 */ + __u64 percpu_offset; /* 0x03b8 */ + __u64 vdso_per_cpu_data; /* 0x03c0 */ + __u64 machine_flags; /* 0x03c8 */ + __u64 gmap; /* 0x03d0 */ + __u8 pad_0x03d8[0x0400-0x03d8]; /* 0x03d8 */ /* br %r1 trampoline */ __u16 br_r1_trampoline; /* 0x0400 */ diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index b198ece2aad6..b8eb99685546 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -266,7 +266,8 @@ static void pcpu_prepare_secondary(struct pcpu *pcpu, int cpu) lc->percpu_offset = __per_cpu_offset[cpu]; lc->kernel_asce = S390_lowcore.kernel_asce; lc->machine_flags = S390_lowcore.machine_flags; - lc->user_timer = lc->system_timer = lc->steal_timer = 0; + lc->user_timer = lc->system_timer = + lc->steal_timer = lc->avg_steal_timer = 0; __ctl_store(lc->cregs_save_area, 0, 15); save_access_regs((unsigned int *) lc->access_regs_save_area); memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, diff --git a/arch/s390/kernel/vtime.c b/arch/s390/kernel/vtime.c index 98f850e00008..a69a0911ed0e 100644 --- a/arch/s390/kernel/vtime.c +++ b/arch/s390/kernel/vtime.c @@ -124,7 +124,7 @@ static void account_system_index_scaled(struct task_struct *p, u64 cputime, */ static int do_account_vtime(struct task_struct *tsk) { - u64 timer, clock, user, guest, system, hardirq, softirq, steal; + u64 timer, clock, user, guest, system, hardirq, softirq; timer = S390_lowcore.last_update_timer; clock = S390_lowcore.last_update_clock; @@ -182,12 +182,6 @@ static int do_account_vtime(struct task_struct *tsk) if (softirq) account_system_index_scaled(tsk, softirq, CPUTIME_SOFTIRQ); - steal = S390_lowcore.steal_timer; - if ((s64) steal > 0) { - S390_lowcore.steal_timer = 0; - account_steal_time(cputime_to_nsecs(steal)); - } - return virt_timer_forward(user + guest + system + hardirq + softirq); } @@ -213,8 +207,19 @@ void vtime_task_switch(struct task_struct *prev) */ void vtime_flush(struct task_struct *tsk) { + u64 steal, avg_steal; + if (do_account_vtime(tsk)) virt_timer_expire(); + + steal = S390_lowcore.steal_timer; + avg_steal = S390_lowcore.avg_steal_timer / 2; + if ((s64) steal > 0) { + S390_lowcore.steal_timer = 0; + account_steal_time(steal); + avg_steal += steal; + } + S390_lowcore.avg_steal_timer = avg_steal; } /* From 50b7f1b7236bab08ebbbecf90521e84b068d7a17 Mon Sep 17 00:00:00 2001 From: Cornelia Huck Date: Mon, 11 Mar 2019 10:59:53 +0100 Subject: [PATCH 4/6] vfio: ccw: only free cp on final interrupt When we get an interrupt for a channel program, it is not necessarily the final interrupt; for example, the issuing guest may request an intermediate interrupt by specifying the program-controlled-interrupt flag on a ccw. We must not switch the state to idle if the interrupt is not yet final; even more importantly, we must not free the translated channel program if the interrupt is not yet final, or the host can crash during cp rewind. Fixes: e5f84dbaea59 ("vfio: ccw: return I/O results asynchronously") Cc: stable@vger.kernel.org # v4.12+ Reviewed-by: Eric Farman Signed-off-by: Cornelia Huck --- drivers/s390/cio/vfio_ccw_drv.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/s390/cio/vfio_ccw_drv.c b/drivers/s390/cio/vfio_ccw_drv.c index a10cec0e86eb..0b3b9de45c60 100644 --- a/drivers/s390/cio/vfio_ccw_drv.c +++ b/drivers/s390/cio/vfio_ccw_drv.c @@ -72,20 +72,24 @@ static void vfio_ccw_sch_io_todo(struct work_struct *work) { struct vfio_ccw_private *private; struct irb *irb; + bool is_final; private = container_of(work, struct vfio_ccw_private, io_work); irb = &private->irb; + is_final = !(scsw_actl(&irb->scsw) & + (SCSW_ACTL_DEVACT | SCSW_ACTL_SCHACT)); if (scsw_is_solicited(&irb->scsw)) { cp_update_scsw(&private->cp, &irb->scsw); - cp_free(&private->cp); + if (is_final) + cp_free(&private->cp); } memcpy(private->io_region->irb_area, irb, sizeof(*irb)); if (private->io_trigger) eventfd_signal(private->io_trigger, 1); - if (private->mdev) + if (private->mdev && is_final) private->state = VFIO_CCW_STATE_IDLE; } From 0d9c038feff6f834ad9e5d88b66715235ab23ff3 Mon Sep 17 00:00:00 2001 From: Tony Krowiak Date: Mon, 18 Feb 2019 12:01:35 -0500 Subject: [PATCH 5/6] zcrypt: handle AP Info notification from CHSC SEI command The current AP bus implementation periodically polls the AP configuration to detect changes. When the AP configuration is dynamically changed via the SE or an SCLP instruction, the changes will not be reflected to sysfs until the next time the AP configuration is polled. The CHSC architecture provides a Store Event Information (SEI) command to make notification of an AP configuration change. This patch introduces a handler to process notification from the CHSC SEI command by immediately kicking off an AP bus scan-after-event. Signed-off-by: Tony Krowiak Reviewed-by: Halil Pasic Reviewed-by: Sebastian Ott Reviewed-by: Harald Freudenberger Reviewed-by: Cornelia Huck Signed-off-by: Sebastian Ott Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/ap.h | 11 +++++++++++ drivers/s390/cio/chsc.c | 13 +++++++++++++ drivers/s390/crypto/ap_bus.c | 10 ++++++++++ 3 files changed, 34 insertions(+) diff --git a/arch/s390/include/asm/ap.h b/arch/s390/include/asm/ap.h index 1a6a7092d942..e94a0a28b5eb 100644 --- a/arch/s390/include/asm/ap.h +++ b/arch/s390/include/asm/ap.h @@ -360,4 +360,15 @@ static inline struct ap_queue_status ap_dqap(ap_qid_t qid, return reg1; } +/* + * Interface to tell the AP bus code that a configuration + * change has happened. The bus code should at least do + * an ap bus resource rescan. + */ +#if IS_ENABLED(CONFIG_ZCRYPT) +void ap_bus_cfg_chg(void); +#else +static inline void ap_bus_cfg_chg(void){}; +#endif + #endif /* _ASM_S390_AP_H_ */ diff --git a/drivers/s390/cio/chsc.c b/drivers/s390/cio/chsc.c index a0baee25134c..eaf4699be2c5 100644 --- a/drivers/s390/cio/chsc.c +++ b/drivers/s390/cio/chsc.c @@ -24,6 +24,7 @@ #include #include #include +#include #include "css.h" #include "cio.h" @@ -586,6 +587,15 @@ static void chsc_process_sei_scm_avail(struct chsc_sei_nt0_area *sei_area) " failed (rc=%d).\n", ret); } +static void chsc_process_sei_ap_cfg_chg(struct chsc_sei_nt0_area *sei_area) +{ + CIO_CRW_EVENT(3, "chsc: ap config changed\n"); + if (sei_area->rs != 5) + return; + + ap_bus_cfg_chg(); +} + static void chsc_process_sei_nt2(struct chsc_sei_nt2_area *sei_area) { switch (sei_area->cc) { @@ -612,6 +622,9 @@ static void chsc_process_sei_nt0(struct chsc_sei_nt0_area *sei_area) case 2: /* i/o resource accessibility */ chsc_process_sei_res_acc(sei_area); break; + case 3: /* ap config changed */ + chsc_process_sei_ap_cfg_chg(sei_area); + break; case 7: /* channel-path-availability information */ chsc_process_sei_chp_avail(sei_area); break; diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c index 033a1acabf48..1546389d71db 100644 --- a/drivers/s390/crypto/ap_bus.c +++ b/drivers/s390/crypto/ap_bus.c @@ -867,6 +867,16 @@ void ap_bus_force_rescan(void) } EXPORT_SYMBOL(ap_bus_force_rescan); +/* +* A config change has happened, force an ap bus rescan. +*/ +void ap_bus_cfg_chg(void) +{ + AP_DBF(DBF_INFO, "%s config change, forcing bus rescan\n", __func__); + + ap_bus_force_rescan(); +} + /* * hex2bitmap() - parse hex mask string and set bitmap. * Valid strings are "0x012345678" with at least one valid hex number. From b6ffdf27f3d4f1e9af56effe6f86989170d71e95 Mon Sep 17 00:00:00 2001 From: Thomas Richter Date: Mon, 18 Mar 2019 15:50:27 +0100 Subject: [PATCH 6/6] s390/cpumf: Fix warning from check_processor_id Function __hw_perf_event_init() used a CPU variable without ensuring CPU preemption has been disabled. This caused the following warning in the kernel log: [ 7.277085] BUG: using smp_processor_id() in preemptible [00000000] code: cf-csdiag/1892 [ 7.277111] caller is cf_diag_event_init+0x13a/0x338 [ 7.277122] CPU: 10 PID: 1892 Comm: cf-csdiag Not tainted 5.0.0-20190318.rc0.git0.9e1a11e0f602.300.fc29.s390x+debug #1 [ 7.277131] Hardware name: IBM 2964 NC9 712 (LPAR) [ 7.277139] Call Trace: [ 7.277150] ([<000000000011385a>] show_stack+0x82/0xd0) [ 7.277161] [<0000000000b7a71a>] dump_stack+0x92/0xd0 [ 7.277174] [<00000000007b7e9c>] check_preemption_disabled+0xe4/0x100 [ 7.277183] [<00000000001228aa>] cf_diag_event_init+0x13a/0x338 [ 7.277195] [<00000000002cf3aa>] perf_try_init_event+0x72/0xf0 [ 7.277204] [<00000000002d0bba>] perf_event_alloc+0x6fa/0xce0 [ 7.277214] [<00000000002dc4a8>] __s390x_sys_perf_event_open+0x398/0xd50 [ 7.277224] [<0000000000b9e8f0>] system_call+0xdc/0x2d8 [ 7.277233] 2 locks held by cf-csdiag/1892: [ 7.277241] #0: 00000000976f5510 (&sig->cred_guard_mutex){+.+.}, at: __s390x_sys_perf_event_open+0xd2e/0xd50 [ 7.277257] #1: 00000000363b11bd (&pmus_srcu){....}, at: perf_event_alloc+0x52e/0xce0 The variable is now accessed in proper context. Use get_cpu_var()/put_cpu_var() pair to disable preemption during access. As the hardware authorization settings apply to all CPUs, it does not matter which CPU is used to check the authorization setting. Remove the event->count assignment. It is not needed as function perf_event_alloc() allocates memory for the event with kzalloc() and thus count is already set to zero. Fixes: fe5908bccc56 ("s390/cpum_cf_diag: Add support for s390 counter facility diagnostic trace") Signed-off-by: Thomas Richter Reviewed-by: Hendrik Brueckner Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/perf_cpum_cf_diag.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) diff --git a/arch/s390/kernel/perf_cpum_cf_diag.c b/arch/s390/kernel/perf_cpum_cf_diag.c index c6fad208c2fa..b6854812d2ed 100644 --- a/arch/s390/kernel/perf_cpum_cf_diag.c +++ b/arch/s390/kernel/perf_cpum_cf_diag.c @@ -196,23 +196,30 @@ static void cf_diag_perf_event_destroy(struct perf_event *event) */ static int __hw_perf_event_init(struct perf_event *event) { - struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events); struct perf_event_attr *attr = &event->attr; + struct cpu_cf_events *cpuhw; enum cpumf_ctr_set i; int err = 0; - debug_sprintf_event(cf_diag_dbg, 5, - "%s event %p cpu %d authorized %#x\n", __func__, - event, event->cpu, cpuhw->info.auth_ctl); + debug_sprintf_event(cf_diag_dbg, 5, "%s event %p cpu %d\n", __func__, + event, event->cpu); event->hw.config = attr->config; event->hw.config_base = 0; - local64_set(&event->count, 0); - /* Add all authorized counter sets to config_base */ + /* Add all authorized counter sets to config_base. The + * the hardware init function is either called per-cpu or just once + * for all CPUS (event->cpu == -1). This depends on the whether + * counting is started for all CPUs or on a per workload base where + * the perf event moves from one CPU to another CPU. + * Checking the authorization on any CPU is fine as the hardware + * applies the same authorization settings to all CPUs. + */ + cpuhw = &get_cpu_var(cpu_cf_events); for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i) if (cpuhw->info.auth_ctl & cpumf_ctr_ctl[i]) event->hw.config_base |= cpumf_ctr_ctl[i]; + put_cpu_var(cpu_cf_events); /* No authorized counter sets, nothing to count/sample */ if (!event->hw.config_base) {