From a7f82c3641245055412b2b4f859ae55fd29fdffe Mon Sep 17 00:00:00 2001 From: Niklas Schnelle Date: Fri, 9 Apr 2021 14:08:50 +0200 Subject: [PATCH 01/10] s390/pci: rename zpci_configure_device() With zpci_configure_device() now always called on a device that has already been configured on the platform level its name has become misleading. Rename it to zpci_scan_configured_device() to signify that the function now only handles the correct scanning of a newly configured PCI function taking care of the special handling necessary for function 0 and functions parked waiting for a PCI bus that can't be created without first seeing function 0. Reviewed-by: Matthew Rosato Reviewed-by: Pierre Morel Signed-off-by: Niklas Schnelle Signed-off-by: Heiko Carstens --- arch/s390/include/asm/pci.h | 2 +- arch/s390/pci/pci.c | 10 ++++++---- arch/s390/pci/pci_event.c | 2 +- drivers/pci/hotplug/s390_pci_hpc.c | 2 +- 4 files changed, 9 insertions(+), 7 deletions(-) diff --git a/arch/s390/include/asm/pci.h b/arch/s390/include/asm/pci.h index 35c2af9371a9..10b67f8aab99 100644 --- a/arch/s390/include/asm/pci.h +++ b/arch/s390/include/asm/pci.h @@ -204,7 +204,7 @@ extern unsigned int s390_pci_no_rid; struct zpci_dev *zpci_create_device(u32 fid, u32 fh, enum zpci_state state); int zpci_enable_device(struct zpci_dev *); int zpci_disable_device(struct zpci_dev *); -int zpci_configure_device(struct zpci_dev *zdev, u32 fh); +int zpci_scan_configured_device(struct zpci_dev *zdev, u32 fh); int zpci_deconfigure_device(struct zpci_dev *zdev); int zpci_register_ioat(struct zpci_dev *, u8, u64, u64, u64); diff --git a/arch/s390/pci/pci.c b/arch/s390/pci/pci.c index c01b6dbac7cf..b0993e05affe 100644 --- a/arch/s390/pci/pci.c +++ b/arch/s390/pci/pci.c @@ -738,17 +738,19 @@ struct zpci_dev *zpci_create_device(u32 fid, u32 fh, enum zpci_state state) } /** - * zpci_configure_device() - Configure a zpci_dev + * zpci_scan_configured_device() - Scan a freshly configured zpci_dev * @zdev: The zpci_dev to be configured * @fh: The general function handle supplied by the platform * * Given a device in the configuration state Configured, enables, scans and - * adds it to the common code PCI subsystem. If any failure occurs, the - * zpci_dev is left disabled. + * adds it to the common code PCI subsystem if possible. If the PCI device is + * parked because we can not yet create a PCI bus because we have not seen + * function 0, it is ignored but will be scanned once function 0 appears. + * If any failure occurs, the zpci_dev is left disabled. * * Return: 0 on success, or an error code otherwise */ -int zpci_configure_device(struct zpci_dev *zdev, u32 fh) +int zpci_scan_configured_device(struct zpci_dev *zdev, u32 fh) { int rc; diff --git a/arch/s390/pci/pci_event.c b/arch/s390/pci/pci_event.c index 1178b48a66df..8ecc256d27a5 100644 --- a/arch/s390/pci/pci_event.c +++ b/arch/s390/pci/pci_event.c @@ -113,7 +113,7 @@ static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf) break; zdev->state = ZPCI_FN_STATE_CONFIGURED; } - zpci_configure_device(zdev, ccdf->fh); + zpci_scan_configured_device(zdev, ccdf->fh); break; case 0x0302: /* Reserved -> Standby */ if (!zdev) diff --git a/drivers/pci/hotplug/s390_pci_hpc.c b/drivers/pci/hotplug/s390_pci_hpc.c index f8f056be71b7..014868752cd4 100644 --- a/drivers/pci/hotplug/s390_pci_hpc.c +++ b/drivers/pci/hotplug/s390_pci_hpc.c @@ -35,7 +35,7 @@ static int enable_slot(struct hotplug_slot *hotplug_slot) return rc; zdev->state = ZPCI_FN_STATE_CONFIGURED; - return zpci_configure_device(zdev, zdev->fh); + return zpci_scan_configured_device(zdev, zdev->fh); } static int disable_slot(struct hotplug_slot *hotplug_slot) From 0d9cf5d8c5d0bfa144236b5f2aeff02124940c56 Mon Sep 17 00:00:00 2001 From: Niklas Schnelle Date: Fri, 9 Apr 2021 13:51:46 +0200 Subject: [PATCH 02/10] s390/pci: handle stale deconfiguration events The PCIs event with PEC 0x0303 or 0x0304 are a request to deconfigure a PCI function, respectively an indication that it was already deconfigured by the platform. If such an event is queued during boot it may happen that the platform has already adjusted the configuration flag of the relevant function in the CLP List PCI Functions result. In this case we might not have configured the PCI function at all and should thus ignore the event. Note that no locking is necessary as event handling only starts after we have fully initialized the zPCI subsystem and scanned all PCI devices listed in the CLP result. Reviewed-by: Matthew Rosato Acked-by: Pierre Morel Signed-off-by: Niklas Schnelle Signed-off-by: Heiko Carstens --- arch/s390/pci/pci_event.c | 26 ++++++++++++++++++-------- 1 file changed, 18 insertions(+), 8 deletions(-) diff --git a/arch/s390/pci/pci_event.c b/arch/s390/pci/pci_event.c index 8ecc256d27a5..cd447b96b4b1 100644 --- a/arch/s390/pci/pci_event.c +++ b/arch/s390/pci/pci_event.c @@ -76,8 +76,6 @@ void zpci_event_error(void *data) static void zpci_event_hard_deconfigured(struct zpci_dev *zdev, u32 fh) { - enum zpci_state state; - zdev->fh = fh; /* Give the driver a hint that the function is * already unusable. @@ -88,15 +86,12 @@ static void zpci_event_hard_deconfigured(struct zpci_dev *zdev, u32 fh) */ zpci_disable_device(zdev); zdev->state = ZPCI_FN_STATE_STANDBY; - if (!clp_get_state(zdev->fid, &state) && - state == ZPCI_FN_STATE_RESERVED) { - zpci_zdev_put(zdev); - } } static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf) { struct zpci_dev *zdev = get_zdev_by_fid(ccdf->fid); + enum zpci_state state; zpci_err("avail CCDF:\n"); zpci_err_hex(ccdf, sizeof(*ccdf)); @@ -123,13 +118,28 @@ static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf) break; case 0x0303: /* Deconfiguration requested */ if (zdev) { + /* The event may have been queued before we confirgured + * the device. + */ + if (zdev->state != ZPCI_FN_STATE_CONFIGURED) + break; zdev->fh = ccdf->fh; zpci_deconfigure_device(zdev); } break; case 0x0304: /* Configured -> Standby|Reserved */ - if (zdev) - zpci_event_hard_deconfigured(zdev, ccdf->fh); + if (zdev) { + /* The event may have been queued before we confirgured + * the device.: + */ + if (zdev->state == ZPCI_FN_STATE_CONFIGURED) + zpci_event_hard_deconfigured(zdev, ccdf->fh); + /* The 0x0304 event may immediately reserve the device */ + if (!clp_get_state(zdev->fid, &state) && + state == ZPCI_FN_STATE_RESERVED) { + zpci_zdev_put(zdev); + } + } break; case 0x0306: /* 0x308 or 0x302 for multiple devices */ zpci_remove_reserved_devices(); From f53a63667b0b30944462ca0ecb9f9dd5b02a56f1 Mon Sep 17 00:00:00 2001 From: Niklas Schnelle Date: Mon, 26 Apr 2021 15:56:47 +0200 Subject: [PATCH 03/10] s390/configs: enable CONFIG_PCI_IOV All major distributions ship with CONFIG_PCI_IOV=y so let us enable it for our defconfigs as well. Note also that since commit e5794cf1a270 ("s390/pci: create links between PFs and VFs") we enabled proper linking between PFs and their associated VFs so with this commit and its fixes applied we can fully support handling SR-IOV enabled PFs. Signed-off-by: Niklas Schnelle Signed-off-by: Heiko Carstens --- arch/s390/configs/debug_defconfig | 1 + arch/s390/configs/defconfig | 1 + 2 files changed, 2 insertions(+) diff --git a/arch/s390/configs/debug_defconfig b/arch/s390/configs/debug_defconfig index 6422618a4f75..4eff0a3810ca 100644 --- a/arch/s390/configs/debug_defconfig +++ b/arch/s390/configs/debug_defconfig @@ -387,6 +387,7 @@ CONFIG_CGROUP_NET_PRIO=y CONFIG_BPF_JIT=y CONFIG_NET_PKTGEN=m CONFIG_PCI=y +CONFIG_PCI_IOV=y # CONFIG_PCIEASPM is not set CONFIG_PCI_DEBUG=y CONFIG_HOTPLUG_PCI=y diff --git a/arch/s390/configs/defconfig b/arch/s390/configs/defconfig index 371a529546aa..8e8bc93fe047 100644 --- a/arch/s390/configs/defconfig +++ b/arch/s390/configs/defconfig @@ -377,6 +377,7 @@ CONFIG_CGROUP_NET_PRIO=y CONFIG_BPF_JIT=y CONFIG_NET_PKTGEN=m CONFIG_PCI=y +CONFIG_PCI_IOV=y # CONFIG_PCIEASPM is not set CONFIG_HOTPLUG_PCI=y CONFIG_HOTPLUG_PCI_S390=y From 0cceeab5a38d70fae3c2944e77e1d262c74d159b Mon Sep 17 00:00:00 2001 From: Thomas Richter Date: Mon, 26 Apr 2021 14:33:53 +0200 Subject: [PATCH 04/10] s390/cpumf: beautify if-then-else indentation Beautify if-then-else indentation to match coding guideline. Also use shorter pointer notation hwc instead of event->hw. Signed-off-by: Thomas Richter Acked-by : Sumanth Korikkar Signed-off-by: Heiko Carstens --- arch/s390/kernel/perf_cpum_cf.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/arch/s390/kernel/perf_cpum_cf.c b/arch/s390/kernel/perf_cpum_cf.c index b3beef64d3d4..931e9d9b6524 100644 --- a/arch/s390/kernel/perf_cpum_cf.c +++ b/arch/s390/kernel/perf_cpum_cf.c @@ -230,9 +230,7 @@ static int __hw_perf_event_init(struct perf_event *event, unsigned int type) /* No support for kernel space counters only */ } else if (!attr->exclude_kernel && attr->exclude_user) { return -EOPNOTSUPP; - - /* Count user and kernel space */ - } else { + } else { /* Count user and kernel space */ if (ev >= ARRAY_SIZE(cpumf_generic_events_basic)) return -EOPNOTSUPP; ev = cpumf_generic_events_basic[ev]; @@ -402,12 +400,12 @@ static void cpumf_pmu_stop(struct perf_event *event, int flags) */ if (!atomic_dec_return(&cpuhw->ctr_set[hwc->config_base])) ctr_set_stop(&cpuhw->state, hwc->config_base); - event->hw.state |= PERF_HES_STOPPED; + hwc->state |= PERF_HES_STOPPED; } if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) { hw_perf_event_update(event); - event->hw.state |= PERF_HES_UPTODATE; + hwc->state |= PERF_HES_UPTODATE; } } From 1eefa4f4399b74dc7671c4e34c1b1c6244acff22 Mon Sep 17 00:00:00 2001 From: Thomas Richter Date: Mon, 26 Apr 2021 14:38:25 +0200 Subject: [PATCH 05/10] s390/cpumf: move counter set size calculation to common place The function to calculate the size of counter sets is renamed from cf_diag_ctrset_size() to cpum_cf_ctrset_size() and moved to the file containing common functions for the CPU Measurement Counter Facility. No functional change. Signed-off-by: Thomas Richter Acked-by : Sumanth Korikkar Signed-off-by: Heiko Carstens --- arch/s390/include/asm/cpu_mcf.h | 2 + arch/s390/kernel/perf_cpum_cf_common.c | 46 ++++++++++++++++++++++ arch/s390/kernel/perf_cpum_cf_diag.c | 54 ++------------------------ 3 files changed, 52 insertions(+), 50 deletions(-) diff --git a/arch/s390/include/asm/cpu_mcf.h b/arch/s390/include/asm/cpu_mcf.h index 649b9fc60685..3e4cbcb7c4cc 100644 --- a/arch/s390/include/asm/cpu_mcf.h +++ b/arch/s390/include/asm/cpu_mcf.h @@ -123,4 +123,6 @@ static inline int stccm_avail(void) return test_facility(142); } +size_t cpum_cf_ctrset_size(enum cpumf_ctr_set ctrset, + struct cpumf_ctr_info *info); #endif /* _ASM_S390_CPU_MCF_H */ diff --git a/arch/s390/kernel/perf_cpum_cf_common.c b/arch/s390/kernel/perf_cpum_cf_common.c index 3bced89caffb..6d53215c8484 100644 --- a/arch/s390/kernel/perf_cpum_cf_common.c +++ b/arch/s390/kernel/perf_cpum_cf_common.c @@ -170,6 +170,52 @@ static int cpum_cf_offline_cpu(unsigned int cpu) return cpum_cf_setup(cpu, PMC_RELEASE); } +/* Return the maximum possible counter set size (in number of 8 byte counters) + * depending on type and model number. + */ +size_t cpum_cf_ctrset_size(enum cpumf_ctr_set ctrset, + struct cpumf_ctr_info *info) +{ + size_t ctrset_size = 0; + + switch (ctrset) { + case CPUMF_CTR_SET_BASIC: + if (info->cfvn >= 1) + ctrset_size = 6; + break; + case CPUMF_CTR_SET_USER: + if (info->cfvn == 1) + ctrset_size = 6; + else if (info->cfvn >= 3) + ctrset_size = 2; + break; + case CPUMF_CTR_SET_CRYPTO: + if (info->csvn >= 1 && info->csvn <= 5) + ctrset_size = 16; + else if (info->csvn == 6) + ctrset_size = 20; + break; + case CPUMF_CTR_SET_EXT: + if (info->csvn == 1) + ctrset_size = 32; + else if (info->csvn == 2) + ctrset_size = 48; + else if (info->csvn >= 3 && info->csvn <= 5) + ctrset_size = 128; + else if (info->csvn == 6) + ctrset_size = 160; + break; + case CPUMF_CTR_SET_MT_DIAG: + if (info->csvn > 3) + ctrset_size = 48; + break; + case CPUMF_CTR_SET_MAX: + break; + } + + return ctrset_size; +} + static int __init cpum_cf_init(void) { int rc; diff --git a/arch/s390/kernel/perf_cpum_cf_diag.c b/arch/s390/kernel/perf_cpum_cf_diag.c index 2e3e7edbe3a0..08c985c1097c 100644 --- a/arch/s390/kernel/perf_cpum_cf_diag.c +++ b/arch/s390/kernel/perf_cpum_cf_diag.c @@ -316,52 +316,6 @@ static void cf_diag_read(struct perf_event *event) debug_sprintf_event(cf_diag_dbg, 5, "%s event %p\n", __func__, event); } -/* Return the maximum possible counter set size (in number of 8 byte counters) - * depending on type and model number. - */ -static size_t cf_diag_ctrset_size(enum cpumf_ctr_set ctrset, - struct cpumf_ctr_info *info) -{ - size_t ctrset_size = 0; - - switch (ctrset) { - case CPUMF_CTR_SET_BASIC: - if (info->cfvn >= 1) - ctrset_size = 6; - break; - case CPUMF_CTR_SET_USER: - if (info->cfvn == 1) - ctrset_size = 6; - else if (info->cfvn >= 3) - ctrset_size = 2; - break; - case CPUMF_CTR_SET_CRYPTO: - if (info->csvn >= 1 && info->csvn <= 5) - ctrset_size = 16; - else if (info->csvn == 6) - ctrset_size = 20; - break; - case CPUMF_CTR_SET_EXT: - if (info->csvn == 1) - ctrset_size = 32; - else if (info->csvn == 2) - ctrset_size = 48; - else if (info->csvn >= 3 && info->csvn <= 5) - ctrset_size = 128; - else if (info->csvn == 6) - ctrset_size = 160; - break; - case CPUMF_CTR_SET_MT_DIAG: - if (info->csvn > 3) - ctrset_size = 48; - break; - case CPUMF_CTR_SET_MAX: - break; - } - - return ctrset_size; -} - /* Calculate memory needed to store all counter sets together with header and * trailer data. This is independend of the counter set authorization which * can vary depending on the configuration. @@ -372,7 +326,7 @@ static size_t cf_diag_ctrset_maxsize(struct cpumf_ctr_info *info) enum cpumf_ctr_set i; for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i) { - size_t size = cf_diag_ctrset_size(i, info); + size_t size = cpum_cf_ctrset_size(i, info); if (size) max_size += size * sizeof(u64) + @@ -405,7 +359,7 @@ static size_t cf_diag_getctrset(struct cf_ctrset_entry *ctrdata, int ctrset, ctrdata->def = CF_DIAG_CTRSET_DEF; ctrdata->set = ctrset; ctrdata->res1 = 0; - ctrset_size = cf_diag_ctrset_size(ctrset, &cpuhw->info); + ctrset_size = cpum_cf_ctrset_size(ctrset, &cpuhw->info); if (ctrset_size) { /* Save data */ need = ctrset_size * sizeof(u64) + sizeof(*ctrdata); @@ -845,7 +799,7 @@ static void cf_diag_cpu_read(void *parm) if (!(p->sets & cpumf_ctr_ctl[set])) continue; /* Counter set not in list */ - set_size = cf_diag_ctrset_size(set, &cpuhw->info); + set_size = cpum_cf_ctrset_size(set, &cpuhw->info); space = sizeof(csd->data) - csd->used; space = cf_diag_cpuset_read(sp, set, set_size, space); if (space) { @@ -975,7 +929,7 @@ static size_t cf_diag_needspace(unsigned int sets) for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i) { if (!(sets & cpumf_ctr_ctl[i])) continue; - bytes += cf_diag_ctrset_size(i, &cpuhw->info) * sizeof(u64) + + bytes += cpum_cf_ctrset_size(i, &cpuhw->info) * sizeof(u64) + sizeof(((struct s390_ctrset_setdata *)0)->set) + sizeof(((struct s390_ctrset_setdata *)0)->no_cnts); } From b0583ab47788617c6af484a844f13c57d7567c2a Mon Sep 17 00:00:00 2001 From: Thomas Richter Date: Mon, 26 Apr 2021 14:57:16 +0200 Subject: [PATCH 06/10] s390/cpumf: remove call to perf_event_update_userpage The function cpumf_pmu_add and cpumf_pmu_del call function perf_event_update_userpage(). This calls is obsolete, the calls add and delete a counter event. Counter events do not sample data and the event->rb member to access the sampling ring buffer is always NULL. The function perf_event_update_userpage() simply returns in this case. Signed-off-by: Thomas Richter Acked-by : Sumanth Korikkar Signed-off-by: Heiko Carstens --- arch/s390/kernel/perf_cpum_cf.c | 4 ---- 1 file changed, 4 deletions(-) diff --git a/arch/s390/kernel/perf_cpum_cf.c b/arch/s390/kernel/perf_cpum_cf.c index 931e9d9b6524..31a605bcbc6e 100644 --- a/arch/s390/kernel/perf_cpum_cf.c +++ b/arch/s390/kernel/perf_cpum_cf.c @@ -428,8 +428,6 @@ static int cpumf_pmu_add(struct perf_event *event, int flags) if (flags & PERF_EF_START) cpumf_pmu_start(event, PERF_EF_RELOAD); - perf_event_update_userpage(event); - return 0; } @@ -449,8 +447,6 @@ static void cpumf_pmu_del(struct perf_event *event, int flags) */ if (!atomic_read(&cpuhw->ctr_set[event->hw.config_base])) ctr_set_disable(&cpuhw->state, event->hw.config_base); - - perf_event_update_userpage(event); } /* From 2f7484fd73729f89085fe08d683f5a8d9e17fe99 Mon Sep 17 00:00:00 2001 From: Vineeth Vijayan Date: Fri, 23 Apr 2021 12:08:43 +0200 Subject: [PATCH 07/10] s390/cio: remove invalid condition on IO_SCH_UNREG The condition to check the cdev pointer validity on css_sch_device_unregister() is a leftover from the 'commit 8cc0dcfdc1c0 ("s390/cio: remove pm support from ccw bus driver")'. This could lead to a situation, where detaching the device is not happening completely. Remove this invalid condition in the IO_SCH_UNREG case. Link: https://lore.kernel.org/r/20210423100843.2230969-1-vneethv@linux.ibm.com Fixes: 8cc0dcfdc1c0 ("s390/cio: remove pm support from ccw bus driver") Reported-by: Christian Ehrhardt Suggested-by: Christian Ehrhardt Cc: Signed-off-by: Vineeth Vijayan Tested-by: Julian Wiedmann Reviewed-by: Peter Oberparleiter Tested-by: Christian Ehrhardt Signed-off-by: Heiko Carstens --- drivers/s390/cio/device.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/s390/cio/device.c b/drivers/s390/cio/device.c index 3f026021e95e..84f659cafe76 100644 --- a/drivers/s390/cio/device.c +++ b/drivers/s390/cio/device.c @@ -1532,8 +1532,7 @@ static int io_subchannel_sch_event(struct subchannel *sch, int process) switch (action) { case IO_SCH_ORPH_UNREG: case IO_SCH_UNREG: - if (!cdev) - css_sch_device_unregister(sch); + css_sch_device_unregister(sch); break; case IO_SCH_ORPH_ATTACH: case IO_SCH_UNREG_ATTACH: From f5b474decad90719e2a4234f83d97aad19307584 Mon Sep 17 00:00:00 2001 From: Thomas Huth Date: Wed, 28 Apr 2021 10:24:42 +0200 Subject: [PATCH 08/10] s390/configs: change CONFIG_VIRTIO_CONSOLE to "m" In former times, the virtio-console code had to be compiled into the kernel since the old guest virtio transport had some hard de- pendencies. But since the old virtio transport has been removed in commit 7fb2b2d51244 ("s390/virtio: remove the old KVM virtio transport"), we do not have this limitation anymore. Commit bb533ec8bacd ("s390/config: do not select VIRTIO_CONSOLE via Kconfig") then also lifted the hard setting in the Kconfig system, so we can finally switch the CONFIG_VIRTIO_CONSOLE knob to compile this driver as a module now, making it more flexible for the user to only load it if it is really required. Signed-off-by: Thomas Huth Link: https://lore.kernel.org/r/20210428082442.321327-1-thuth@redhat.com Signed-off-by: Christian Borntraeger Signed-off-by: Heiko Carstens --- arch/s390/configs/debug_defconfig | 2 +- arch/s390/configs/defconfig | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/s390/configs/debug_defconfig b/arch/s390/configs/debug_defconfig index 4eff0a3810ca..86afcc6b56bf 100644 --- a/arch/s390/configs/debug_defconfig +++ b/arch/s390/configs/debug_defconfig @@ -549,7 +549,7 @@ CONFIG_INPUT_EVDEV=y # CONFIG_INPUT_MOUSE is not set # CONFIG_SERIO is not set CONFIG_LEGACY_PTY_COUNT=0 -CONFIG_VIRTIO_CONSOLE=y +CONFIG_VIRTIO_CONSOLE=m CONFIG_HW_RANDOM_VIRTIO=m CONFIG_RAW_DRIVER=m CONFIG_HANGCHECK_TIMER=m diff --git a/arch/s390/configs/defconfig b/arch/s390/configs/defconfig index 8e8bc93fe047..71b49ea5b058 100644 --- a/arch/s390/configs/defconfig +++ b/arch/s390/configs/defconfig @@ -541,7 +541,7 @@ CONFIG_INPUT_EVDEV=y # CONFIG_INPUT_MOUSE is not set # CONFIG_SERIO is not set CONFIG_LEGACY_PTY_COUNT=0 -CONFIG_VIRTIO_CONSOLE=y +CONFIG_VIRTIO_CONSOLE=m CONFIG_HW_RANDOM_VIRTIO=m CONFIG_RAW_DRIVER=m CONFIG_HANGCHECK_TIMER=m From bae1cd368c45d1127e054e90305d585dbc8b3b46 Mon Sep 17 00:00:00 2001 From: Sven Schnelle Date: Thu, 29 Apr 2021 11:14:51 +0200 Subject: [PATCH 09/10] s390/entry: add support for syscall stack randomization This adds support for adding a random offset to the stack while handling syscalls. The patch uses get_tod_clock_fast() as this is considered good enough and has much less performance penalty compared to using get_random_int(). The patch also adds randomization in pgm_check_handler() as the sigreturn/rt_sigreturn system calls might be called from there. Signed-off-by: Sven Schnelle Link: https://lore.kernel.org/r/20210429091451.1062594-1-svens@linux.ibm.com Signed-off-by: Heiko Carstens --- arch/s390/Kconfig | 1 + arch/s390/include/asm/entry-common.h | 10 ++++++++++ arch/s390/kernel/syscall.c | 1 + arch/s390/kernel/traps.c | 2 ++ 4 files changed, 14 insertions(+) diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index c1ff874e6c2e..1900428ce557 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -137,6 +137,7 @@ config S390 select HAVE_ARCH_JUMP_LABEL_RELATIVE select HAVE_ARCH_KASAN select HAVE_ARCH_KASAN_VMALLOC + select HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET select HAVE_ARCH_SECCOMP_FILTER select HAVE_ARCH_SOFT_DIRTY select HAVE_ARCH_TRACEHOOK diff --git a/arch/s390/include/asm/entry-common.h b/arch/s390/include/asm/entry-common.h index 9cceb26ed63f..baa8005090c3 100644 --- a/arch/s390/include/asm/entry-common.h +++ b/arch/s390/include/asm/entry-common.h @@ -4,9 +4,11 @@ #include #include +#include #include #include #include +#include #include #define ARCH_EXIT_TO_USER_MODE_WORK (_TIF_GUARDED_STORAGE | _TIF_PER_TRAP) @@ -48,6 +50,14 @@ static __always_inline void arch_exit_to_user_mode(void) #define arch_exit_to_user_mode arch_exit_to_user_mode +static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs, + unsigned long ti_work) +{ + choose_random_kstack_offset(get_tod_clock_fast() & 0xff); +} + +#define arch_exit_to_user_mode_prepare arch_exit_to_user_mode_prepare + static inline bool on_thread_stack(void) { return !(((unsigned long)(current->stack) ^ current_stack_pointer()) & ~(THREAD_SIZE - 1)); diff --git a/arch/s390/kernel/syscall.c b/arch/s390/kernel/syscall.c index bc8e650e377d..4e5cc7d2364e 100644 --- a/arch/s390/kernel/syscall.c +++ b/arch/s390/kernel/syscall.c @@ -142,6 +142,7 @@ void do_syscall(struct pt_regs *regs) void noinstr __do_syscall(struct pt_regs *regs, int per_trap) { + add_random_kstack_offset(); enter_from_user_mode(regs); memcpy(®s->gprs[8], S390_lowcore.save_area_sync, 8 * sizeof(unsigned long)); diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c index 63021d484626..8dd23c703718 100644 --- a/arch/s390/kernel/traps.c +++ b/arch/s390/kernel/traps.c @@ -17,6 +17,7 @@ #include "asm/ptrace.h" #include #include +#include #include #include #include @@ -301,6 +302,7 @@ void noinstr __do_pgm_check(struct pt_regs *regs) unsigned int trapnr, syscall_redirect = 0; irqentry_state_t state; + add_random_kstack_offset(); regs->int_code = *(u32 *)&S390_lowcore.pgm_ilc; regs->int_parm_long = S390_lowcore.trans_exc_code; From b208108638c4bd3215792415944467c36f5dfd97 Mon Sep 17 00:00:00 2001 From: David Hildenbrand Date: Mon, 3 May 2021 14:12:44 +0200 Subject: [PATCH 10/10] s390: fix detection of vector enhancements facility 1 vs. vector packed decimal facility The PoP documents: 134: The vector packed decimal facility is installed in the z/Architecture architectural mode. When bit 134 is one, bit 129 is also one. 135: The vector enhancements facility 1 is installed in the z/Architecture architectural mode. When bit 135 is one, bit 129 is also one. Looks like we confuse the vector enhancements facility 1 ("EXT") with the Vector packed decimal facility ("BCD"). Let's fix the facility checks. Detected while working on QEMU/tcg z14 support and only unlocking the vector enhancements facility 1, but not the vector packed decimal facility. Fixes: 2583b848cad0 ("s390: report new vector facilities") Cc: Vasily Gorbik Signed-off-by: David Hildenbrand Reviewed-by: Christian Borntraeger Reviewed-by: Cornelia Huck Reviewed-by: Janosch Frank Link: https://lore.kernel.org/r/20210503121244.25232-1-david@redhat.com Signed-off-by: Heiko Carstens --- arch/s390/kernel/setup.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index 72134f9f6ff5..5aab59ad5688 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -937,9 +937,9 @@ static int __init setup_hwcaps(void) if (MACHINE_HAS_VX) { elf_hwcap |= HWCAP_S390_VXRS; if (test_facility(134)) - elf_hwcap |= HWCAP_S390_VXRS_EXT; - if (test_facility(135)) elf_hwcap |= HWCAP_S390_VXRS_BCD; + if (test_facility(135)) + elf_hwcap |= HWCAP_S390_VXRS_EXT; if (test_facility(148)) elf_hwcap |= HWCAP_S390_VXRS_EXT2; if (test_facility(152))