- 22 Mar, 2018 40 commits
-
-
Gao Feng authored
[ Upstream commit c4836742 ] Because sysctl_tcp_adv_win_scale could be changed any time, so there is one race in tcp_win_from_space. For example, 1.sysctl_tcp_adv_win_scale<=0 (sysctl_tcp_adv_win_scale is negative now) 2.space>>(-sysctl_tcp_adv_win_scale) (sysctl_tcp_adv_win_scale is postive now) As a result, tcp_win_from_space returns 0. It is unexpected. Certainly if the compiler put the sysctl_tcp_adv_win_scale into one register firstly, then use the register directly, it would be ok. But we could not depend on the compiler behavior. Signed-off-by:
Gao Feng <fgao@ikuai8.com> Signed-off-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
Sasha Levin <alexander.levin@microsoft.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Akinobu Mita authored
[ Upstream commit 81261359 ] When running the spi-loopback-test with slower clock rate like 10 KHz, the test for 251 bytes transfer was failed. This failure triggered an spi-omap2-mcspi's error message "DMA RX last word empty". This message means that PIO for reading the remaining bytes due to the DMA transfer length reduction is failed. This problem can be fixed by polling OMAP2_MCSPI_CHSTAT_RXS bit in channel status register to wait until the receive buffer register is filled. Cc: Mark Brown <broonie@kernel.org> Signed-off-by:
Akinobu Mita <akinobu.mita@gmail.com> Signed-off-by:
Mark Brown <broonie@kernel.org> Signed-off-by:
-
Kuninori Morimoto authored
[ Upstream commit 6b8530cc ] R-Car Datasheet is indicating "SSICR.CKDV = 000 is invalid when SSIWSR.WS_MODE = 1 or SSIWSR.CONT = 1". Current driver will set CONT, thus, we shouldn't use CKDV = 000. This patch fixup it. Reported-by:
Hiroyuki Yokoyama <hiroyuki.yokoyama.vx@renesas.com> Signed-off-by:
Kuninori Morimoto <kuninori.morimoto.gx@renesas.com> Tested-by:
-
Long Li authored
[ Upstream commit 414428c5 ] A PCI_EJECT message can arrive at the same time we are calling pci_scan_child_bus() in the workqueue for the previous PCI_BUS_RELATIONS message or in create_root_hv_pci_bus(). In this case we could potentially modify the bus from multiple places. Properly lock the bus access. Thanks Dexuan Cui <decui@microsoft.com> for pointing out the race condition in create_root_hv_pci_bus(). Reported-by:
Xiaofeng Wang <xiaofwan@redhat.com> Signed-off-by:
Long Li <longli@microsoft.com> Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Acked-by:
K. Y. Srinivasan <kys@microsoft.com> Signed-off-by:
-
Long Li authored
[ Upstream commit d3a78d8b ] hv_pci_devices_present() is called in hv_pci_remove() when we remove a PCI device from the host, e.g., by disabling SR-IOV on a device. In hv_pci_remove(), the bus is already removed before the call, so we don't need to rescan the bus in the workqueue scheduled from hv_pci_devices_present(). By introducing bus state hv_pcibus_removed, we can avoid this situation. Reported-by:
-
Davide Caratti authored
[ Upstream commit add641e7 ] after act_csum computes the checksum on skbs carrying GSO TCP/UDP packets, subsequent segmentation fails because skb_needs_check(skb, true) returns true. Because of that, skb_warn_bad_offload() is invoked and the following message is displayed: WARNING: CPU: 3 PID: 28 at net/core/dev.c:2553 skb_warn_bad_offload+0xf0/0xfd <...> [<ffffffff8171f486>] skb_warn_bad_offload+0xf0/0xfd [<ffffffff8161304c>] __skb_gso_segment+0xec/0x110 [<ffffffff8161340d>] validate_xmit_skb+0x12d/0x2b0 [<ffffffff816135d2>] validate_xmit_skb_list+0x42/0x70 [<ffffffff8163c560>] sch_direct_xmit+0xd0/0x1b0 [<ffffffff8163c760>] __qdisc_run+0x120/0x270 [<ffffffff81613b3d>] __dev_queue_xmit+0x23d/0x690 [<ffffffff81613fa0>] dev_queue_xmit+0x10/0x20 Since GSO is able to compute checksum on individual segments of such skbs, we can simply skip mangling the packet. Signed-off-by:
Davide Caratti <dcaratti@redhat.com> Signed-off-by:
-
Javier Martinez Canillas authored
[ Upstream commit cf5cd9d4 ] The driver doesn't have a struct of_device_id table but supported devices are registered via Device Trees. This is working on the assumption that a I2C device registered via OF will always match a legacy I2C device ID and that the MODALIAS reported will always be of the form i2c:<device>. But this could change in the future so the correct approach is to have an OF device ID table if the devices are registered via OF. The compatible strings don't have a vendor prefix because that's how it's used currently, and changing this will be a Device Tree ABI break. Signed-off-by:
Javier Martinez Canillas <javier@osg.samsung.com> Signed-off-by:
Dmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by:
-
Tom Hromatka authored
[ Upstream commit 01070427 ] On systems with a large number of CPUs, running sysrq-<q> can cause watchdog timeouts. There are two slow sections of code in the sysrq-<q> path in timer_list.c. 1. print_active_timers() - This function is called by print_cpu() and contains a slow goto loop. On a machine with hundreds of CPUs, this loop took approximately 100ms for the first CPU in a NUMA node. (Subsequent CPUs in the same node ran much quicker.) The total time to print all of the CPUs is ultimately long enough to trigger the soft lockup watchdog. 2. print_tickdevice() - This function outputs a large amount of textual information. This function also took approximately 100ms per CPU. Since sysrq-<q> is not a performance critical path, there should be no harm in touching the nmi watchdog in both slow sections above. Touching it in just one location was insufficient on systems with hundreds of CPUs as occasional timeouts were still observed during testing. This issue was observed on an Oracle T7 machine with 128 CPUs, but I anticipate it may affect other systems with similarly large numbers of CPUs. Signed-off-by:
Tom Hromatka <tom.hromatka@oracle.com> Reviewed-by:
Rob Gardner <rob.gardner@oracle.com> Signed-off-by:
John Stultz <john.stultz@linaro.org> Signed-off-by:
-
David Engraf authored
[ Upstream commit 1b8955bc ] The scheduler clock framework may not use the correct timeout for the clock wrap. This happens when a new clock driver calls sched_clock_register() after the kernel called sched_clock_postinit(). In this case the clock wrap timeout is too long thus sched_clock_poll() is called too late and the clock already wrapped. On my ARM system the scheduler was no longer scheduling any other task than the idle task because the sched_clock() wrapped. Signed-off-by:
David Engraf <david.engraf@sysgo.com> Signed-off-by:
-
Janusz Krzysztofik authored
[ Upstream commit 54449af0 ] After changes to v4l2_clk API introduced in v4.1 by commits a37462b9 '[media] V4L: remove clock name from v4l2_clk API' and 4f528afc '[media] V4L: add CCF support to the v4l2_clk API', ov6650 sensor stopped responding because v4l2_clk_get(), still called with depreciated V4L2 clock name "mclk", started to return respective CCF clock instead of the V4l2 one registered by soc_camera. Fix it by calling v4l2_clk_get() with NULL clock name. Created and tested on Amstrad Delta against Linux-4.7-rc3 with omap1_camera fixes. Signed-off-by:
Janusz Krzysztofik <jmkrzyszt@gmail.com> Signed-off-by:
Guennadi Liakhovetski <g.liakhovetski@gmx.de> Signed-off-by:
Mauro Carvalho Chehab <mchehab@s-opensource.com> Signed-off-by:
-
Brian King authored
[ Upstream commit 66a0d59c ] Following a command abort or device reset, ipr's EH handlers wait for the commands getting aborted to get sent back from the adapter prior to returning from the EH handler. This fixes up some cases where the completion handler was not getting called, which would have resulted in the EH thread waiting until it timed out, greatly extending EH time. Signed-off-by:
Brian King <brking@linux.vnet.ibm.com> Reviewed-by:
Wendy Xiong <wenxiong@linux.vnet.ibm.com> Tested-by:
Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by:
-
Satish Kharat authored
[ Upstream commit 7ef539c8 ] Fixing the IO stats update (Active IOs and IO completion) to prevent "Number of Active IOs" from becoming negative in the fnistats output. Signed-off-by:
Satish Kharat <satishkh@cisco.com> Signed-off-by:
Sesidhar Baddela <sebaddel@cisco.com> Signed-off-by:
-
Andy Lutomirski authored
[ Upstream commit 23b2a4dd ] The x86 smpboot trampoline expects initial_page_table to have the GDT mapped. If the GDT ends up in a virtually mapped per-cpu page, then it won't be in the page tables at all until perc-pu areas are set up. The result will be a triple fault the first time that the CPU attempts to access the GDT after LGDT loads the perc-pu GDT. This appears to be an old bug, but somehow the GDT fixmap rework is triggering it. This seems to have something to do with the memory layout. Signed-off-by:
Andy Lutomirski <luto@kernel.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Garnier <thgarnie@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-efi@vger.kernel.org Link: http://lkml.kernel.org/r/a553264a5972c6a86f9b5caac237470a0c74a720.1490218061.git.luto@kernel.orgSigned-off-by:
Ingo Molnar <mingo@kernel.org> Signed-off-by:
-
Anton Sviridenko authored
[ Upstream commit 6e4c8480 ] Fixes warning that appears in dmesg after closing V4L2 userspace application that plays video from the display device (first device from V4L2 device nodes provided by solo, usually /dev/video0 when no other V4L2 devices are present). Encoder device nodes are not affected. Can be reproduced by starting and closing ffplay -f video4linux2 /dev/video0 [ 8130.281251] ------------[ cut here ]------------ [ 8130.281256] WARNING: CPU: 1 PID: 20414 at drivers/media/v4l2-core/videobuf2-core.c:1651 __vb2_queue_cancel+0x14b/0x230 [ 8130.281257] Modules linked in: ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat solo6x10 x86_pkg_temp_thermal vboxpci(O) vboxnetadp(O) vboxnetflt(O) vboxdrv(O) [ 8130.281264] CPU: 1 PID: 20414 Comm: ffplay Tainted: G O 4.10.0-gentoo #1 [ 8130.281264] Hardware name: ASUS All Series/B85M-E, BIOS 2301 03/30/2015 [ 8130.281265] Call Trace: [ 8130.281267] dump_stack+0x4f/0x72 [ 8130.281270] __warn+0xc7/0xf0 [ 8130.281271] warn_slowpath_null+0x18/0x20 [ 8130.281272] __vb2_queue_cancel+0x14b/0x230 [ 8130.281273] vb2_core_streamoff+0x23/0x90 [ 8130.281275] vb2_streamoff+0x24/0x50 [ 8130.281276] vb2_ioctl_streamoff+0x3d/0x50 [ 8130.281278] v4l_streamoff+0x15/0x20 [ 8130.281279] __video_do_ioctl+0x25e/0x2f0 [ 8130.281280] video_usercopy+0x279/0x520 [ 8130.281282] ? v4l_enum_fmt+0x1330/0x1330 [ 8130.281285] ? unmap_region+0xdf/0x110 [ 8130.281285] video_ioctl2+0x10/0x20 [ 8130.281286] v4l2_ioctl+0xce/0xe0 [ 8130.281289] do_vfs_ioctl+0x8b/0x5b0 [ 8130.281290] ? __fget+0x72/0xa0 [ 8130.281291] SyS_ioctl+0x74/0x80 [ 8130.281294] entry_SYSCALL_64_fastpath+0x13/0x94 [ 8130.281295] RIP: 0033:0x7ff86fee6b27 [ 8130.281296] RSP: 002b:00007ffe467f6a08 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 8130.281297] RAX: ffffffffffffffda RBX: 00000000d1a4d788 RCX: 00007ff86fee6b27 [ 8130.281297] RDX: 00007ffe467f6a14 RSI: 0000000040045613 RDI: 0000000000000006 [ 8130.281298] RBP: 000000000373f8d0 R08: 00000000ffffffff R09: 00007ff860001140 [ 8130.281298] R10: 0000000000000243 R11: 0000000000000246 R12: 0000000000000000 [ 8130.281299] R13: 00000000000000a0 R14: 00007ffe467f6530 R15: 0000000001f32228 [ 8130.281300] ---[ end trace 00695dc96be646e7 ]--- Signed-off-by:
Anton Sviridenko <anton@corp.bluecherry.net> Signed-off-by:
Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by:
-
Rob Herring authored
[ Upstream commit bcf54d53 ] If the length of the modalias is greater than the buffer size, then the modalias is truncated. However the untruncated length is returned which will cause an error. Fix this to return the truncated length. If an error in the case was desired, then then we should just return -ENOMEM. The reality is no device will ever have 4KB of compatible strings to hit this case. Signed-off-by:
Rob Herring <robh@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Signed-off-by:
-
Andreas Pape authored
[ Upstream commit a3a5129e ] Consider the following situation which has been found in a test setup: Gateway B has claimed client C and gateway A has the same backbone network as B. C sends a broad- or multicast to B and directly after this packet decides to send another packet to A due to a better TQ value. B will forward the broad-/multicast into the backbone as it is the responsible gw and after that A will claim C as it has been chosen by C as the best gateway. If it now happens that A claims C before it has received the broad-/multicast forwarded by B (due to backbone topology or due to some delay in B when forwarding the packet) we get a critical situation: in the current code A will immediately unclaim C when receiving the multicast due to the roaming client scenario although the position of C has not changed in the mesh. If this happens the multi-/broadcast forwarded by B will be sent back into the mesh by A and we have looping packets until one of the gateways claims C again. In order to prevent this, unclaiming of a client due to the roaming client scenario is only done after a certain time is expired after the last claim of the client. 100 ms are used here, which should be slow enough for big backbones and slow gateways but fast enough not to break the roaming client use case. Acked-by:
Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by:
Andreas Pape <apape@phoenixcontact.com> [sven@narfation.org: fix conflicts with current version] Signed-off-by:
Sven Eckelmann <sven@narfation.org> Signed-off-by:
-
Johan Hovold authored
[ Upstream commit ca260ece ] Make sure to check the number of endpoints to avoid dereferencing a NULL-pointer or accessing memory beyond the endpoint array should a malicious device lack the expected endpoints. Fixes: a1030e92 ("[PATCH] zd1211rw: Convert installer CDROM device into WLAN device") Cc: Daniel Drake <dsd@gentoo.org> Signed-off-by:
Johan Hovold <johan@kernel.org> Signed-off-by:
Kalle Valo <kvalo@codeaurora.org> Signed-off-by:
-
Heiko Carstens authored
[ Upstream commit 4fd4dd8b ] Use MACHINE_FLAG_TOPOLOGY instead of MACHINE_HAS_TOPOLOGY when clearing the bit that indicates if the machine provides topology information (and if it should be used). Currently works anyway. Fixes: 68cc795d ("s390/topology: make "topology=off" parameter work") Signed-off-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by:
-
Mintz, Yuval authored
[ Upstream commit e50728ef ] The link information exists only on the leading hwfn, but some of its derivatives [e.g., min/max rate] need to be configured for each hwfn. When re-basing the VF link view, use the leading hwfn information as basis for all existing hwfns to allow said configurations to stick. Signed-off-by:
Yuval Mintz <Yuval.Mintz@cavium.com> Signed-off-by:
-
Linus Walleij authored
[ Upstream commit c2a736b6 ] The moxart interrupt line flags were not respected in previous driver: instead of assigning them per-consumer, a fixes mask was set in the controller. With the migration to a standard Faraday driver we need to set up and handle the consumer flags correctly. Also remove the Moxart-specific flags when switching to using real consumer flags. Extend the register window to 0x100 bytes as we may have a few more registers in there and it doesn't hurt. Tested-by:
Jonas Jensen <jonas.jensen@gmail.com> Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Signed-off-by:
Olof Johansson <olof@lixom.net> Signed-off-by:
-
Andrey Vagin authored
[ Upstream commit 88997e42 ] wanted_features is a set of features which have to be enabled if a hardware allows that. Currently when a vlan device is created, its wanted_features is set to current features of its base device. The problem is that the base device can get new features and they are not propagated to vlan-s of this device. If we look at bonding devices, they doesn't have this problem and this patch suggests to fix this issue by the same way how it works for bonding devices. We meet this problem, when we try to create a vlan device over a bonding device. When a system are booting, real devices require time to be initialized, so bonding devices created without slaves, then vlan devices are created and only then ethernet devices are added to the bonding device. As a result we have vlan devices with disabled scatter-gather. * create a bonding device $ ip link add bond0 type bond $ ethtool -k bond0 | grep scatter scatter-gather: off tx-scatter-gather: off [requested on] tx-scatter-gather-fraglist: off [requested on] * create a vlan device $ ip link add link bond0 name bond0.10 type vlan id 10 $ ethtool -k bond0.10 | grep scatter scatter-gather: off tx-scatter-gather: off tx-scatter-gather-fraglist: off * Add a slave device to bond0 $ ip link set dev eth0 master bond0 And now we can see that the bond0 device has got the scatter-gather feature, but the bond0.10 hasn't got it. [root@laptop linux-task-diag]# ethtool -k bond0 | grep scatter scatter-gather: on tx-scatter-gather: on tx-scatter-gather-fraglist: on [root@laptop linux-task-diag]# ethtool -k bond0.10 | grep scatter scatter-gather: off tx-scatter-gather: off tx-scatter-gather-fraglist: off With this patch the vlan device will get all new features from the bonding device. Here is a call trace how features which are set in this patch reach dev->wanted_features. register_netdevice vlan_dev_init ... dev->hw_features = NETIF_F_HW_CSUM | NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_GSO_SOFTWARE | NETIF_F_HIGHDMA | NETIF_F_SCTP_CRC | NETIF_F_ALL_FCOE; dev->features |= dev->hw_features; ... dev->wanted_features = dev->features & dev->hw_features; __netdev_update_features(dev); vlan_dev_fix_features ... Cc: Alexey Kuznetsov <kuznet@virtuozzo.com> Cc: Patrick McHardy <kaber@trash.net> Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by:
Andrei Vagin <avagin@openvz.org> Signed-off-by:
-
Tomasz Kramkowski authored
[ Upstream commit c3883fe0 ] This patch fixes an issue in drivers/hid/hid-input.c where values outside of the logical range are not clamped when "null state" bit of the input control is not set. This was discussed on the lists [1] and this change stems from the fact due to the ambiguity of the HID specification it might be appropriate to follow Microsoft's own interpretation of the specification. As noted in Microsoft's documentation [2] in the section titled "Required HID usages for digitizers" it is noted that values reported outside the logical range "will be considered as invalid data and the value will be changed to the nearest boundary value (logical min/max)." This patch fixes an issue where the (1292:4745) Innomedia INNEX GENESIS/ATARI reports out of range values for its X and Y axis of the DPad which, due to the null state bit being unset, are forwarded to userspace as is. Now these values will get clamped to the logical range before being forwarded to userspace. This device was also used to test this patch. This patch expands on commit 3f375270 ("HID: reject input outside logical range only if null state is set"). [1]: http://lkml.kernel.org/r/20170307131036.GA853@gaia.local [2]: https://msdn.microsoft.com/en-us/library/windows/hardware/dn672278(v=vs.85).aspSigned-off-by:
Tomasz Kramkowski <tk@the-tk.com> Acked-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com> Signed-off-by:
Jiri Kosina <jkosina@suse.cz> Signed-off-by:
-
Kefeng Wang authored
[ Upstream commit 70946723 ] On old perf, when using 'perf probe -d' to delete an inexistent event, it returns errno, eg, -bash-4.3# perf probe -d xxx || echo $? Info: Event "*:xxx" does not exist. Error: Failed to delete events. 255 But now perf_del_probe_events() will always set ret = 0, different from previous del_perf_probe_events(). After this, it returns errno again, eg, -bash-4.3# ./perf probe -d xxx || echo $? "xxx" does not hit any event. Error: Failed to delete events. 254 And it is more appropriate to return -ENOENT instead of -EPERM. Signed-off-by:
Kefeng Wang <wangkefeng.wang@huawei.com> Acked-by:
Masami Hiramatsu <mhiramat@kernel.org> Cc: Hanjun Guo <guohanjun@huawei.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Wang Nan <wangnan0@huawei.com> Fixes: dddc7ee3 ("perf probe: Fix an error when deleting probes successfully") Link: http://lkml.kernel.org/r/1489738592-61011-1-git-send-email-wangkefeng.wang@huawei.comSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by:
-
Ravi Bangoria authored
[ Upstream commit f0a30dca ] '*ntevs' contains number of elements present in 'tevs' array. If there are no elements in array, 'tevs2' can be directly assigned to 'tevs' without allocating more space. So the condition should be '*ntevs == 0' not 'ntevs == 0'. Signed-off-by:
Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Acked-by:
-
Arvind Yadav authored
[ Upstream commit 0348aaa3 ] dss_init_ports() is not handling return errors from dpi_init_port() and sdi_init_port(). It is also always returning 0 currently which results in part of error handling code in dss_bind() being unused. Fix dss_init_ports() to handle return errors from dpi_init_port() and sdi_init_port(). Signed-off-by:
Arvind Yadav <arvind.yadav.cs@gmail.com> Cc: tomi.valkeinen@ti.com [b.zolnierkie: fail early on errors, minor fixups] Signed-off-by:
Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Signed-off-by:
-
Yazen Ghannam authored
[ Upstream commit 5204bf17 ] When the MCA banks in __mcheck_cpu_init_generic() are polled for leftover errors logged during boot or from the previous boot, its required to have CPU features detected sufficiently so that the reading out and handling of those early errors is done correctly. If those features are not available, the decoding may miss some information and get incomplete errors logged. For example, on SMCA systems the MCA_IPID and MCA_SYND registers are not logged and MCA_ADDR is not masked appropriately. To cure that, do a subset of the basic feature detection early while the rest happens in its usual place in __mcheck_cpu_init_vendor(). Signed-off-by:
Yazen Ghannam <Yazen.Ghannam@amd.com> Cc: Tony Luck <tony.luck@intel.com> Cc: linux-edac <linux-edac@vger.kernel.org> Cc: x86-ml <x86@kernel.org> Link: http://lkml.kernel.org/r/1489599055-20756-1-git-send-email-Yazen.Ghannam@amd.com [ Massage commit message and simplify. ] Signed-off-by:
Borislav Petkov <bp@suse.de> Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
-
Nik Unger authored
[ Upstream commit 5080f39e ] I recently reported on the netem list that iperf network benchmarks show unexpected results when a bandwidth throttling rate has been configured for netem. Specifically: 1) The measured link bandwidth *increases* when a higher delay is added 2) The measured link bandwidth appears higher than the specified limit 3) The measured link bandwidth for the same very slow settings varies significantly across machines The issue can be reproduced by using tc to configure netem with a 512kbit rate and various (none, 1us, 50ms, 100ms, 200ms) delays on a veth pair between network namespaces, and then using iperf (or any other network benchmarking tool) to test throughput. Complete detailed instructions are in the original email chain here: https://lists.linuxfoundation.org/pipermail/netem/2017-February/001672.html There appear to be two underlying bugs causing these effects: - The first issue causes long delays when the rate is slow and no delay is configured (e.g., "rate 512kbit"). This is because SKBs are not orphaned when no delay is configured, so orphaning does not occur until *after* the rate-induced delay has been applied. For this reason, adding a tiny delay (e.g., "rate 512kbit delay 1us") dramatically increases the measured bandwidth. - The second issue is that rate-induced delays are not correctly applied, allowing SKB delays to occur in parallel. The indended approach is to compute the delay for an SKB and to add this delay to the end of the current queue. However, the code does not detect existing SKBs in the queue due to improperly testing sch->q.qlen, which is nonzero even when packets exist only in the rbtree. Consequently, new SKBs do not wait for the current queue to empty. When packet delays vary significantly (e.g., if packet sizes are different), then this also causes unintended reordering. I modified the code to expect a delay (and orphan the SKB) when a rate is configured. I also added some defensive tests that correctly find the latest scheduled delivery time, even if it is (unexpectedly) for a packet in sch->q. I have tested these changes on the latest kernel (4.11.0-rc1+) and the iperf / ping test results are as expected. Signed-off-by:
Nik Unger <njunger@uwaterloo.ca> Signed-off-by:
Stephen Hemminger <stephen@networkplumber.org> Signed-off-by:
-
Steve Lin authored
[ Upstream commit 2f771399 ] Allows the BCMA version of the bgmac driver to obtain MAC address from the device tree. If no MAC address is specified there, then the previous behavior (obtaining MAC address from SPROM) is used. Signed-off-by:
Steve Lin <steven.lin1@broadcom.com> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Acked-by:
Jon Mason <jon.mason@broadcom.com> Signed-off-by:
-
Stefan Wahren authored
[ Upstream commit bdd3c254 ] Currently bcm2835_defconfig has CMA disabled which makes the HDMI output on a Raspberry Pi 1 stop working during boot: fb: switching to vc4drmfb from simple Console: switching to colour dummy device 80x30 [drm] Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0 [drm] Supports vblank timestamp caching Rev 2 (21.10.2013). [drm] Driver supports precise vblank timestamp query. vc4-drm soc:gpu: failed to allocate buffer with size 9216000 vc4-drm soc:gpu: Failed to set initial hw configuration. So enable CMA and DMA_CMA in bcm2835_defconfig. Signed-off-by:
Stefan Wahren <stefan.wahren@i2se.com> Fixes: 4400d9ac ("ARM: bcm2835: Enable the VC4 graphics driver in the defconfig") Reviewed-by:
Eric Anholt <eric@anholt.net> Signed-off-by:
-
Oliver Neukum authored
[ Upstream commit c4ba329c ] There is a small window during which the an URB may remain active after disconnect has returned. If in that case already freed memory may be accessed and executed. The fix is to poison the URB befotre the work is flushed. Signed-off-by:
Oliver Neukum <oneukum@suse.com> Signed-off-by:
-
Mohammed Shafi Shajakhan authored
[ Upstream commit a28f6f27 ] Fetch target operating channel during potential radar detection when the interface is just brought up, but no channel is assigned from userspace. In this scenario rx_channel may not be having a valid pointer hence fetch the target operating channel to avoid warnings as below which can be triggered by the commands with DFS testing over longer run comamnds: iw wlan1 set type mesh ifconfig wlan1 up (valid tgt_oper_chan only) iw wlan1 cac trigger freq 5260 HT20 (valid rx_channel, tgt_oper_chan) iw wlan1 cac trigger freq 5280 HT20 iw wlan1 cac trigger freq 5300 HT20 Once the CAC expires, current channel context will be removed and we are only left with the fallback option of using 'target operating channel' Firmware and driver log: ath: phy1: DFS: radar found on freq=5300: id=1, pri=1125, count=5, count_false=4 ath: phy1: DFS: radar found on freq=5260: id=5, pri=3151, count=6, count_false=11 ath: phy1: DFS: radar found on freq=5280: id=1, pri=1351, count=6, count_false=4 ath: phy1: DFS: radar found on freq=5300: id=1, pri=1125, count=5, count_false=4 ath10k_pci 0001:01:00.0: failed to derive channel for radar pulse, treating as radar ath10k_pci 0001:01:00.0: failed to derive channel for radar pulse, treating as radar Call trace: WARNING: CPU: 1 PID: 2145 at backports-20161201-3.14.77-9ab3068/net/wireless/chan.c:265 cfg80211_set_dfs_state+0x3c/0x88 [cfg80211]() Workqueue: phy1 ieee80211_dfs_radar_detected_work [mac80211] [<c0320770>] (warn_slowpath_null) from [<bf79b90c>] (cfg80211_set_dfs_state+0x3c/0x88 [cfg80211]) [<bf79b90c>] (cfg80211_set_dfs_state [cfg80211]) from [<bf79697c>] (cfg80211_radar_event+0xc4/0x140 [cfg80211]) [<bf79697c>] (cfg80211_radar_event [cfg80211]) from [<bf83c058>] (ieee80211_dfs_radar_detected_work+0xa8/0xb4 [mac80211]) [<bf83c058>] (ieee80211_dfs_radar_detected_work [mac80211]) from [<c0339518>] (process_one_work+0x298/0x4a4) Signed-off-by:
Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com> Signed-off-by:
Kalle Valo <kvalo@qca.qualcomm.com> Signed-off-by:
-
Mohammed Shafi Shajakhan authored
[ Upstream commit ca07baab ] If DFS is not enabled in hostapd (ieee80211h=0) DFS channels shall not be available for use even though the hardware may have the capability to support DFS. With this configuration (DFS disabled in hostapd) trying to bring up ath10k device in DFS channel for AP mode fails and trying to simulate DFS in ath10k debugfs results in a warning in cfg80211 complaining invalid channel and this should be avoided in the driver itself rather than false propogating RADAR detection to mac80211/cfg80211. Fix this by checking for the first vif 'is_started' state(should work for client mode as well) as all the vifs shall be configured for the same channel sys/kernel/debug/ieee80211/phy1/ath10k# echo 1 > dfs_simulate_radar WARNING: at net/wireless/chan.c:265 cfg80211_radar_event+0x24/0x60 Workqueue: phy0 ieee80211_dfs_radar_detected_work [mac80211] [<c022f2d4>] (warn_slowpath_null) from [<bf72dab8>] (cfg80211_radar_event+0x24/0x60 [cfg80211]) [<bf72dab8>] (cfg80211_radar_event [cfg80211]) from [<bf7813e0>] (ieee80211_dfs_radar_detected_work+0x94/0xa0 [mac80211]) [<bf7813e0>] (ieee80211_dfs_radar_detected_work [mac80211]) from [<c0242320>] (process_one_work+0x20c/0x32c) WARNING: at net/wireless/nl80211.c:2488 nl80211_get_mpath+0x13c/0x4cc Workqueue: phy0 ieee80211_dfs_radar_detected_work [mac80211] [<c022f2d4>] (warn_slowpath_null) from [<bf72dab8>] (cfg80211_radar_event+0x24/0x60 [cfg80211]) [<bf72dab8>] (cfg80211_radar_event [cfg80211]) from [<bf7813e0>] (ieee80211_dfs_radar_detected_work+0x94/0xa0 [mac80211]) [<bf7813e0>] (ieee80211_dfs_radar_detected_work [mac80211]) from [<c0242320>] (process_one_work+0x20c/0x32c) Signed-off-by:
-
Chris Wilson authored
[ Upstream commit 608b2050 ] On vblank instant-off systems, we can get into a situation where the cost of enabling and disabling the vblank IRQ around a drmWaitVblank query dominates. And with the advent of even deeper hardware sleep state, touching registers becomes ever more expensive. However, we know that if the user wants the current vblank counter, they are also very likely to immediately queue a vblank wait and so we can keep the interrupt around and only turn it off if we have no further vblank requests queued within the interrupt interval. After vblank event delivery, this patch adds a shadow of one vblank where the interrupt is kept alive for the user to query and queue another vblank event. Similarly, if the user is using blocking drmWaitVblanks, the interrupt will be disabled on the IRQ following the wait completion. However, if the user is simply querying the current vblank counter and timestamp, the interrupt will be disabled after every IRQ and the user will enabled it again on the first query following the IRQ. v2: Mario Kleiner - After testing this, one more thing that would make sense is to move the disable block at the end of drm_handle_vblank() instead of at the top. Turns out that if high precision timestaming is disabled or doesn't work for some reason (as can be simulated by echo 0 > /sys/module/drm/parameters/timestamp_precision_usec), then with your delayed disable code at its current place, the vblank counter won't increment anymore at all for instant queries, ie. with your other "instant query" patches. Clients which repeatedly query the counter and wait for it to progress will simply hang, spinning in an endless query loop. There's that comment in vblank_disable_and_save: "* Skip this step if there isn't any high precision timestamp * available. In that case we can't account for this and just * hope for the best. */ With the disable happening after leading edge of vblank (== hw counter increment already happened) but before the vblank counter/timestamp handling in drm_handle_vblank, that step is needed to keep the counter progressing, so skipping it is bad. Now without high precision timestamping support, a kms driver must not set dev->vblank_disable_immediate = true, as this would cause problems for clients, so this shouldn't matter, but it would be good to still make this robust against a future kms driver which might have unreliable high precision timestamping, e.g., high precision timestamping that intermittently doesn't work. v3: Patch before coffee needs extra coffee. Testcase: igt/kms_vblank Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Michel Dänzer <michel@daenzer.net> Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Cc: Dave Airlie <airlied@redhat.com>, Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20170315204027.20160-1-chris@chris-wilson.co.ukSigned-off-by:
-
Iyappan Subramanian authored
[ Upstream commit 0a0400c3 ] This patch fixes Rx checksum validation logic and adds NETIF_F_RXCSUM flag. Signed-off-by:
Iyappan Subramanian <isubramanian@apm.com> Signed-off-by:
-
Quan Nguyen authored
[ Upstream commit 11623fce ] This patch fixes the wrong logical OR operation by changing it to bit-wise OR operation. Fixes: 3bb502f8 ("drivers: net: xgene: fix statistics counters race condition") Signed-off-by:
Quan Nguyen <qnguyen@apm.com> Signed-off-by:
-
Quan Nguyen authored
[ Upstream commit 4b72436d ] This patches fixes a typo in the argument to xgene_enet_wr_mdio_csr(). Signed-off-by:
-
Quan Nguyen authored
[ Upstream commit e026e700 ] This patch fixes the hardware checksum settings by properly program the classifier. Otherwise, packet may be received with checksum error on X-Gene1 SoC. Signed-off-by:
-
Al Cooper authored
[ Upstream commit 3c51b9c7 ] Some Host Controller hardware blocks, like the OHCI, EHCI and SDIO controllers, have hardware blocks that are not capable of doing 64 bit DMA. These host controllers fail on boards with >3GB of memory because the memory above 3GB is located physically >= 0x100000000 and can only be accessed using 64 DMA. The way Linux is currently configured for BRCMSTB systems, the memory given to drivers for DMA through functions like dma_alloc_coherent() comes from CMA memory and CMA memory is taken from the top of physical memory. When these drivers get a DMA buffer with an address >=0x100000000, they end up dropping the upper 32 bit of the address causing the hardware to DMA to incorrect memory, typically BMEM (custom memory carveout). This issue was discovered on a BCM97449SSV_DDR4 system with 4GB or memory. The fix is to enable CONFIG_ZONE_DMA. On ARM systems this makes sure that all DMA memory is located within the first 32 bits of address space. Signed-off-by:
Al Cooper <alcooperx@gmail.com> Signed-off-by:
-
Stephane Eranian authored
[ Upstream commit 88b897a3 ] This patch significantly improves the execution time of perf_event__synthesize_mmap_events() when running perf record on systems where processes have lots of threads. It just happens that cat /proc/pid/maps support uses a O(N^2) algorithm to generate each map line in the maps file. If you have 1000 threads, then you have necessarily 1000 stacks. For each vma, you need to check if it corresponds to a thread's stack. With a large number of threads, this can take a very long time. I have seen latencies >> 10mn. As of today, perf does not use the fact that a mapping is a stack, therefore we can work around the issue by using /proc/pid/tasks/pid/maps. This entry does not try to map a vma to stack and is thus much faster with no loss of functonality. The proc-map-timeout logic is kept in case users still want some upper limit. In V2, we fix the file path from /proc/pid/tasks/pid/maps to actual /proc/pid/task/pid/maps, tasks -> task. Thanks Arnaldo for catching this. Committer note: This problem seems to have been elliminated in the kernel since commit : b18cb64e ("fs/proc: Stop trying to report thread stacks"). Signed-off-by:
Stephane Eranian <eranian@google.com> Acked-by:
Jiri Olsa <jolsa@redhat.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20170315135059.GC2177@redhat.com Link: http://lkml.kernel.org/r/1489598233-25586-1-git-send-email-eranian@google.comSigned-off-by:
-
Lihong Yang authored
[ Upstream commit c271dd6c ] Currently ethtool -e will error out with a X722 interface as its EEPROM has a scope limit at offset 0x5B9FFF. This patch fixes the issue by setting the EEPROM length to the scope limit to avoid NVM read failure beyond that. Change-ID: I0b7d4dd6c7f2a57cace438af5dffa0f44c229372 Signed-off-by:
Lihong Yang <lihong.yang@intel.com> Tested-by:
Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by:
-
