- 16 Dec, 2013 36 commits
-
-
Sebastian Ott authored
When calling set_{on,off}line of a ccwgroup device driver we hold the module reference of the owner. This is pretty useless - we don't want to prevent module unloading but driver unbinding. Use the driver core's device_lock instead. Reviewed-by:Peter Oberparleiter <oberpar@linux.vnet.ibm.com> Signed-off-by:
Sebastian Ott <sebott@linux.vnet.ibm.com> Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com>
-
Sebastian Ott authored
When calling set_{on,off}line of a ccw device driver we hold the module reference of the owner. This is pretty useless - we don't want to prevent module unloading but driver unbinding. Use the driver core's device_lock instead. Reviewed-by: -
Sebastian Ott authored
Make sure that access to the online member of a ccw device is guarded by the ccwlock. Reported-by:
Cornelia Huck <cornelia.huck@de.ibm.com> Reviewed-by:
-
Hendrik Brueckner authored
Add the PERF_CPUM_SF_FULL_BLOCKS flag to process only sample-data-blocks that have the block-full-indicator bit set. Sample-data-blocks that are partially filled are discarded. Use this flag if the sampling buffer is likely to be shared among perf events that use different sampling modes. In such environments, flushing sample-data-blocks that are not completely filled, might cause invalid-data-formats. Setting PERF_CPUM_SF_FULL_BLOCKS prevents potentially invalid sampling data to be processed but, in contrast, also discards valid samples in partially filled sample-data-blocks. Note that sample-data-blocks might not become full for small sampling frequencies or for workload that is scheduled for tiny intervals. To sample with the PERF_CPUM_SF_FULL_BLOCKS flag, set the perf->attr.config1 to 0x0004. For example: perf record -e cpum_sf/config=0xB000,config1=0x0004/ Signed-off-by:
Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Signed-off-by:
-
Hendrik Brueckner authored
Also support the diagnostic-sampling function in addition to the basic-sampling function. Diagnostic-sampling data entries contain hardware model specific sampling data and additional programs are required to analyze the data. To deliver diagnostic-sampling, as well, as basis-sampling data entries to user space, introduce support for sampling "raw data". If this particular perf sampling type (PERF_SAMPLE_RAW) is used, sampling data entries are copied to user space. External programs can then analyze these data. Signed-off-by:
-
Hendrik Brueckner authored
Introduce the perf_exclude_event() function to filter perf samples according to event->attr.exclude_* settings. During event initialization, reset event exclude settings that are not supported. Signed-off-by:
-
Hendrik Brueckner authored
The host-program-parameter (hpp) value of basic sample-data-entries designates a SIE control block that is set by the LPP instruction in sie64a(). Non-zero values indicate guest samples, a value of zero indicates a host sample. For perf samples, host and guest samples are distinguished using particular PERF_MISC_* flags. The perf layer calls perf_misc_flags() to set the flags based on the pt_regs content. For each sample-data-entry, the cpum_sf PMU creates a pt_regs structure with the sample-data information. An additional flag structure is added to easily distinguish between host and guest samples. Signed-off-by:
-
Hendrik Brueckner authored
The trailer entry contains a timestamp of the time when the sample-data-block became full. The timestamp specifies a TOD (time-of-day) value in either the STCK or STCKE format. Provide a helper function to return the TOD value depending on the setting of time format indicator. Signed-off-by:
-
Hendrik Brueckner authored
Ensure to reset the sample-data-block full indicator and the overflow counter at the same time. This must be done atomically because the sampling hardware is still active while full sample-data-block is processed. Signed-off-by:
-
Hendrik Brueckner authored
Improve the sampling buffer allocation and add a function to reallocate and increase the sampling buffer structure. The number of allocated buffer elements (sample-data-blocks) are accounted. You can control the minimum and maximum number these sample-data-blocks through the cpum_sfb_size kernel parameter. The number hardware sample overflows (if any) are also accounted and stored per perf event. During the PMU disable/enable calls, the accumulated overflow counter is analyzed and, if necessary, the sampling buffer is dynamically increased. Signed-off-by:
-
Sebastian Ott authored
HW, FW and Linux support is in a better shape now - let's reenable pci bus probing per default. Acked-by:
Gerald Schaefer <gerald.schaefer@de.ibm.com> Signed-off-by:
-
Sebastian Ott authored
allocated_pages sometimes are increased even if s390_dma_alloc fails also this value is never decreased even if s390_dma_free is called. This patch fixes these bugs. Also remove the atomic64_t casts (the members are already of this type). Reviewed-by:
-
Sebastian Ott authored
If we receive a notification that a pci function became unavailable we clean up by removing the pci device. This can confuse the driver since the function is already unaccessible. Improve this situation by setting an appropriate error_state. Reviewed-by:
-
Sebastian Ott authored
If we remove a pci bus after receiving a hotplug notification we need to check if the bus is actually present (creation of the pci bus during an earlier notification may have been failed). Reviewed-by:
-
Sebastian Ott authored
Initialization and scanning of the pci bus is omitted on older machines without pci support or if pci=off was specified. Remember the fact that we ran without pci support and prevent further bus scans during resume from hibernate or after receiving hotplug notifications. Reported-by:
Stefan Haberland <stefan.haberland@de.ibm.com> Reviewed-by:
-
Hendrik Brueckner authored
Register a service level handler to report information about available CPU-Measurement facilities. Signed-off-by:
-
Hendrik Brueckner authored
Introduce reserve/release functions to share the sampling facility between perf and oprofile. Also improve error handling for the sampling facility support in perf. Signed-off-by:
-
Hendrik Brueckner authored
The cpum_cf (counter facility) PMU does not support sampling events. With cpum_sf (sampling facility), a PMU for sampling CPU cycles is available. Make cpum_sf the "default" PMU for PERF_COUNT_HW_CPU_CYCLES sampling events but use the more precise cpum_cf PMU for non-sampling events. Signed-off-by:
-
Hendrik Brueckner authored
Introduce a perf PMU, "cpum_sf", to support the CPU-Measurement Sampling Facility. You can control the sampling facility through this perf PMU interfaces. Perf sampling events are created for hardware samples. For details about the CPU-Measurement Sampling Facility, see "The Load-Program-Parameter and the CPU-Measurement Facilities" (SA23-2260). Signed-off-by:
-
Hendrik Brueckner authored
Provide PMU event attributes for supported counters and export their symbolic names to the sysfs "events" directory. See the /sys/devices/cpum_cf/events/ directory for a list of available counters. Note that you might require counter set authorizations for the LPAR to use them. Signed-off-by:
-
Hendrik Brueckner authored
Extract and move the oprofile hwsampler data structures and interfaces to the cpu_mf.h header file which contains common interface definitions for the various CPU-measurement facilities. This change is necessary for a new perf PMU. Few interface names have been revised to fit to the latest CPU-measurement facilities documentation. Also declare the data structures as __packed and correct checkpatch findings. Signed-off-by:
-
Hendrik Brueckner authored
Add SCLP console detect functions to encapsulate detection of SCLP console capabilities, for example, VT220 support. Reuse the sclp_send/receive masks that were stored by the most recent sclp_set_event_mask() call to prevent unnecessary SCLP calls. Signed-off-by:
Michael Holzheu <holzheu@linux.vnet.ibm.com> Reviewed-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
-
Hendrik Brueckner authored
Add a sccb pointer parameter to *_detect() functions instead of accessing the global sccb_early variable directly. Signed-off-by:
-
Hendrik Brueckner authored
Replace early_read_info_sccb and use sccb_early instead. Also saves some memory. Signed-off-by:
-
Hendrik Brueckner authored
The early sclp detect functions gather the available SCLP facility information. The sclp_early_read_info_sccb_valid indicates whether the early sclp request was valid. However, one external reference to it checks for particular sclp facility bits and this should be sufficient. Another occurance is in the sclp_get_ipl_info() function that is called later. Because all information are available at the early stage, save the ipl information when detecting the sclp facilities. Hence, no more checks for sclp_early_read_info_sccb_valid are required. Signed-off-by:
-
Martin Schwidefsky authored
The return code of the __put_user call to store the rt_sigreturn system call to the user stack if not properly checked, the err variable is only checked before to the __put_user. Use an if statement instead. Signed-off-by: -
Heiko Carstens authored
Remove the embedded struct cpu from struct pcpu and replace it with a pointer instead. The struct cpu now gets allocated when a new cpu gets detected. The size of the pcpu_devices array (NR_CPUS * sizeof(struct pcpu)) gets reduced by nearly 120KB. Signed-off-by:
-
Martin Schwidefsky authored
The deactivation and freeing of the tty view of the 3270 device can race with a tty3270_update invocation via the update timer. To fix this move the del_timer_sync call for the update timer from tty3270_free_view to tty3270_free prior to the tty3270_free_screen call. Signed-off-by: -
Martin Schwidefsky authored
It is less expensive to update control registers 0 and 2 with two individual stctg/lctlg instructions as with a single one that spans control register 0, 1 and 2. Signed-off-by: -
Martin Schwidefsky authored
The user_enable_single_step() and user_disable_sindle_step() functions are always called on the inferior, never for the currently active process. Remove the unnecessary check for the current process and the update_cr_regs() call from the enable/disable functions. Signed-off-by: -
Heiko Carstens authored
If the per cpu ec_mask bit of the receiving cpu is already set there is no need to send an ipi, since a different cpu has already sent an ipi and the receiving cpu has not yet executed the external call ipi handler. Signed-off-by:
-
Peter Oberparleiter authored
Move scheduling of a subchannel scan to those instances where new devices may actually have become available. This reduces unnecessary scan work in case devices were added to the blacklist. Signed-off-by:
-
Peter Oberparleiter authored
Subchannel looping function for_each_subchannel_staged() allocates a subchannel-ID-bitmap to efficiently iterate over the list of known and unknown subchannels. Since this function is also used to iterate over known-subchannels only, optimize that case by not requiring the ID-bitmap allocation and falling back to simple bus_for_each_dev() looping. Signed-off-by:
-
Peter Oberparleiter authored
The CIO layer scans for newly available I/O devices by performing a scan of available subchannels using the Store Subchannel (STSCH) instruction. Performing too many STSCH instructions in a tight loop can cause high Hypervisor overhead which can negatively impact the performance of the virtual machine as a whole. A subchannel scan is triggered for example during a hardware event that indicates that a channel path has become available. It is also triggered by the DASD device driver for each device that is set online. This patch reduces the number of STSCH instructions being performed by delaying the start of the actual subchannel scan by 1 second. Multiple scan requests that are scheduled during this time will be merged into a single scan loop. The trade-off consists of a short delay that is introduced between the time that the event is processed and a newly available device becoming usable. This delay should be acceptable since it only affects devices that have not been in use before. Signed-off-by:
-
Peter Oberparleiter authored
The CIO layer scans for newly available I/O devices by performing a scan of available subchannels using the Store Subchannel (STSCH) instruction. This processing can take a significant amount of time during which no other task can run on the same CPU (unless CONFIG_PREEMPT has been enabled). As a result, scheduling latencies for other tasks are increased noticeably, especially on a single-CPU system. Fix this problem by explicitly allowing other tasks to be scheduled each time a subchannel has been processed. Signed-off-by:
-
Peter Oberparleiter authored
The CIO layer processes hardware events that indicate that a channel path has become available by performing a scan of available subchannels using the Store Subchannel (STSCH) instruction. Performing too many STSCH instructions in a tight loop can cause high Hypervisor overhead which can negatively impact the performance of the virtual machine as a whole. This patch reduces the number of STSCH instructions performed while processing a resource accessibility event and while varying a CHPID online. In both cases, Linux first performs a STSCH instruction on each unused subchannel to see if the subchannel has become available. If the STSCH instruction indicates that the subchannel is available, a full evaluation of this subchannel is scheduled. Since the full evaluation includes performing a STSCH instruction, the initial STSCH is unnecessary and can be removed. Signed-off-by:
-
- 15 Dec, 2013 4 commits
-
-
Linus Torvalds authored
-
Matias Bjorling authored
For NUMA systems, initializing the blk-mq layer and using per node hctx. We initialize submit queues to 1, while blk-mq nr_hw_queues is initialized to the number of NUMA nodes. This makes the null_init_hctx function overwrite memory outside of what it allocated. In my case it lead to writing garbage into struct request_queue's mq_map. Signed-off-by:
Matias Bjorling <m@bjorling.me> Cc: Jens Axboe <axboe@kernel.dk> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org>
-
Sergey Senozhatsky authored
Since commit ec39f64b ("drm/radeon/dpm: Convert to use devm_hwmon_register_with_groups") radeon_hwmon_init() is using hwmon_device_register_with_groups(), which sets `rdev' as a device private driver_data, while hwmon_attributes_visible() and radeon_hwmon_show_temp_thresh() are still waiting for `drm_device'. Fix them by using dev_get_drvdata(), in order to avoid this oops: BUG: unable to handle kernel paging request at 0000000000001e28 IP: [<ffffffffa02ae8b4>] hwmon_attributes_visible+0x18/0x3d [radeon] PGD 15057e067 PUD 151a8e067 PMD 0 Oops: 0000 [#1] PREEMPT SMP Call Trace: internal_create_group+0x114/0x1d9 sysfs_create_group+0xe/0x10 sysfs_create_groups+0x22/0x5f device_add+0x34f/0x501 device_register+0x15/0x18 hwmon_device_register_with_groups+0xb5/0xed radeon_hwmon_init+0x56/0x7c [radeon] radeon_pm_init+0x134/0x7e5 [radeon] radeon_modeset_init+0x75f/0x8ed [radeon] radeon_driver_load_kms+0xc6/0x187 [radeon] drm_dev_register+0xf9/0x1b4 [drm] drm_get_pci_dev+0x98/0x129 [drm] radeon_pci_probe+0xa3/0xac [radeon] pci_device_probe+0x6e/0xcf driver_probe_device+0x98/0x1c4 __driver_attach+0x5c/0x7e bus_for_each_dev+0x7b/0x85 driver_attach+0x19/0x1b bus_add_driver+0x104/0x1ce driver_register+0x89/0xc5 __pci_register_driver+0x58/0x5b drm_pci_init+0x86/0xea [drm] radeon_init+0x97/0x1000 [radeon] do_one_initcall+0x7f/0x117 load_module+0x1583/0x1bb4 SyS_init_module+0xa0/0xaf Signed-off-by:
Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Alexander Deucher <Alexander.Deucher@amd.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/davem/netLinus Torvalds authored
Pull networking fixes from David Miller: 1) Revert CHECKSUM_COMPLETE optimization in pskb_trim_rcsum(), I can't figure out why it breaks things. 2) Fix comparison in netfilter ipset's hash_netnet4_data_equal(), it was basically doing "x == x", from Dave Jones. 3) Freescale FEC driver was DMA mapping the wrong number of bytes, from Sebastian Siewior. 4) Blackhole and prohibit routes in ipv6 were not doing the right thing because their ->input and ->output methods were not being assigned correctly. Now they behave properly like their ipv4 counterparts. From Kamala R. 5) Several drivers advertise the NETIF_F_FRAGLIST capability, but really do not support this feature and will send garbage packets if fed fraglist SKBs. From Eric Dumazet. 6) Fix long standing user triggerable BUG_ON over loopback in RDS protocol stack, from Venkat Venkatsubra. 7) Several not so common code paths can potentially try to invoke packet scheduler actions that might be NULL without checking. Shore things up by either 1) defining a method as mandatory and erroring on registration if that method is NULL 2) defininig a method as optional and the registration function hooks up a default implementation when NULL is seen. From Jamal Hadi Salim. 8) Fix fragment detection in xen-natback driver, from Paul Durrant. 9) Kill dangling enter_memory_pressure method in cg_proto ops, from Eric W Biederman. 10) SKBs that traverse namespaces should have their local_df cleared, from Hannes Frederic Sowa. 11) IOCB file position is not being updated by macvtap_aio_read() and tun_chr_aio_read(). From Zhi Yong Wu. 12) Don't free virtio_net netdev before releasing all of the NAPI instances. From Andrey Vagin. 13) Procfs entry leak in xt_hashlimit, from Sergey Popovich. 14) IPv6 routes that are no cached routes should not count against the garbage collection limits. We had this almost right, but were missing handling addrconf generated routes properly. From Hannes Frederic Sowa. 15) fib{4,6}_rule_suppress() have to consider potentially seeing NULL route info when they are called, from Stefan Tomanek. 16) TUN and MACVTAP have had truncated packet signalling for some time, fix from Jason Wang. 17) Fix use after frrr in __udp4_lib_rcv(), from Eric Dumazet. 18) xen-netback does not interpret the NAPI budget properly for TX work, fix from Paul Durrant. * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (132 commits) igb: Fix for issue where values could be too high for udelay function. i40e: fix null dereference xen-netback: fix gso_prefix check net: make neigh_priv_len in struct net_device 16bit instead of 8bit drivers: net: cpsw: fix for cpsw crash when build as modules xen-netback: napi: don't prematurely request a tx event xen-netback: napi: fix abuse of budget sch_tbf: use do_div() for 64-bit divide udp: ipv4: must add synchronization in udp_sk_rx_dst_set() net:fec: remove duplicate lines in comment about errata ERR006358 Revert "8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature" 8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature xen-netback: make sure skb linear area covers checksum field net: smc91x: Fix device tree based configuration so it's usable udp: ipv4: fix potential use after free in udp_v4_early_demux() macvtap: signal truncated packets tun: unbreak truncated packet signalling net: sched: htb: fix the calculation of quantum net: sched: tbf: fix the calculation of max_size micrel: add support for KSZ8041RNLI ...
-
