- 05 Mar, 2019 40 commits
-
-
BOUGH CHEN authored
commit e30be063 upstream. Commit 18094430 ("mmc: sdhci-esdhc-imx: add ADMA Length Mismatch errata fix") involve the fix of ERR004536, but the fix is incorrect. Double confirm with IC, need to clear the bit 7 of register 0x6c rather than set this bit 7. Here is the definition of bit 7 of 0x6c: 0: enable the new IC fix for ERR004536 1: do not use the IC fix, keep the same as before Find this issue on i.MX845s-evk board when enable CMDQ, and let system in heavy loading. root@imx8mmevk:~# dd if=/dev/mmcblk2 of=/dev/null bs=1M & root@imx8mmevk:~# memtester 1000M > /dev/zero & root@imx8mmevk:~# [ 139.897220] mmc2: cqhci: timeout for tag 16 [ 139.901417] mmc2: cqhci: ============ CQHCI REGISTER DUMP =========== [ 139.907862] mmc2: cqhci: Caps: 0x0000310a | Version: 0x00000510 [ 139.914311] mmc2: cqhci: Config: 0x00001001 | Control: 0x00000000 [ 139.920753] mmc2: cqhci: Int stat: 0x00000000 | Int enab: 0x00000006 [ 139.927193] mmc2: cqhci: Int sig: 0x00000006 | Int Coal: 0x00000000 [ 139.933634] mmc2: cqhci: TDL base: 0x7809c000 | TDL up32: 0x00000000 [ 139.940073] mmc2: cqhci: Doorbell: 0x00030000 | TCN: 0x00000000 [ 139.946518] mmc2: cqhci: Dev queue: 0x00010000 | Dev Pend: 0x00010000 [ 139.952967] mmc2: cqhci: Task clr: 0x00000000 | SSC1: 0x00011000 [ 139.959411] mmc2: cqhci: SSC2: 0x00000001 | DCMD rsp: 0x00000000 [ 139.965857] mmc2: cqhci: RED mask: 0xfdf9a080 | TERRI: 0x00000000 [ 139.972308] mmc2: cqhci: Resp idx: 0x0000002e | Resp arg: 0x00000900 [ 139.978761] mmc2: sdhci: ============ SDHCI REGISTER DUMP =========== [ 139.985214] mmc2: sdhci: Sys addr: 0xb2c19000 | Version: 0x00000002 [ 139.991669] mmc2: sdhci: Blk size: 0x00000200 | Blk cnt: 0x00000400 [ 139.998127] mmc2: sdhci: Argument: 0x40110400 | Trn mode: 0x00000033 [ 140.004618] mmc2: sdhci: Present: 0x01088a8f | Host ctl: 0x00000030 [ 140.011113] mmc2: sdhci: Power: 0x00000002 | Blk gap: 0x00000080 [ 140.017583] mmc2: sdhci: Wake-up: 0x00000008 | Clock: 0x0000000f [ 140.024039] mmc2: sdhci: Timeout: 0x0000008f | Int stat: 0x00000000 [ 140.030497] mmc2: sdhci: Int enab: 0x107f4000 | Sig enab: 0x107f4000 [ 140.036972] mmc2: sdhci: AC12 err: 0x00000000 | Slot int: 0x00000502 [ 140.043426] mmc2: sdhci: Caps: 0x07eb0000 | Caps_1: 0x8000b407 [ 140.049867] mmc2: sdhci: Cmd: 0x00002c1a | Max curr: 0x00ffffff [ 140.056314] mmc2: sdhci: Resp[0]: 0x00000900 | Resp[1]: 0xffffffff [ 140.062755] mmc2: sdhci: Resp[2]: 0x328f5903 | Resp[3]: 0x00d00f00 [ 140.069195] mmc2: sdhci: Host ctl2: 0x00000008 [ 140.073640] mmc2: sdhci: ADMA Err: 0x00000007 | ADMA Ptr: 0x7809c108 [ 140.080079] mmc2: sdhci: ============================================ [ 140.086662] mmc2: running CQE recovery Fixes: 18094430 ("mmc: sdhci-esdhc-imx: add ADMA Length Mismatch errata fix") Signed-off-by:
Haibo Chen <haibo.chen@nxp.com> Cc: stable@vger.kernel.org Signed-off-by:
Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Alamy Liu authored
commit d07e9fad upstream. Free up the allocated memory in the case of error return The value of mmc_host->cqe_enabled stays 'false'. Thus, cqhci_disable (mmc_cqe_ops->cqe_disable) won't be called to free the memory. Also, cqhci_disable() seems to be designed to disable and free all resources, not suitable to handle this corner case. Fixes: a4080225 ("mmc: cqhci: support for command queue enabled host") Signed-off-by:
Alamy Liu <alamy.liu@gmail.com> Acked-by:
Adrian Hunter <adrian.hunter@intel.com> Cc: stable@vger.kernel.org Signed-off-by:
-
Alamy Liu authored
commit 27ec9dc1 upstream. There is not enough space being allocated when DCMD is disabled. CQE_DCMD is not necessary to be enabled when CQE is enabled. (Software could halt CQE to send command) In the case that CQE_DCMD is not enabled, it still needs to allocate space for data transfer. For instance: CQE_DCMD is enabled: 31 slots space (one slot used by DCMD) CQE_DCMD is disabled: 32 slots space Fixes: a4080225 ("mmc: cqhci: support for command queue enabled host") Signed-off-by:
-
Ritesh Harjani authored
commit e5723f95 upstream. In case of CQHCI, mrq->cmd may be NULL for data requests (non DCMD). In such case mmc_should_fail_request is directly dereferencing mrq->cmd while cmd is NULL. Fix this by checking for mrq->cmd pointer. Fixes: 72a5af55 ("mmc: core: Add support for handling CQE requests") Signed-off-by:
Ritesh Harjani <riteshh@codeaurora.org> Cc: stable@vger.kernel.org Signed-off-by:
-
Takeshi Saito authored
commit 5603731a upstream. In R-Car Gen2 or later, the maximum number of transfer blocks are changed from 0xFFFF to 0xFFFFFFFF. Therefore, Block Count Register should use iowrite32(). If another system (U-boot, Hypervisor OS, etc) uses bit[31:16], this value will not be cleared. So, SD/MMC card initialization fails. So, check for the bigger register and use apropriate write. Also, mark the register as extended on Gen2. Signed-off-by:
Takeshi Saito <takeshi.saito.xv@renesas.com> [wsa: use max_blk_count in if(), add Gen2, update commit message] Signed-off-by:
Wolfram Sang <wsa+renesas@sang-engineering.com> Cc: stable@kernel.org Reviewed-by:
Simon Horman <horms+renesas@verge.net.au> [Ulf: Fixed build error] Signed-off-by:
-
Sergei Shtylyov authored
commit 5c27ff5d upstream. I have encountered an interrupt storm during the eMMC chip probing (and the chip finally didn't get detected). It turned out that U-Boot left the DMAC interrupts enabled while the Linux driver didn't use those. The SDHI driver's interrupt handler somehow assumes that, even if an SDIO interrupt didn't happen, it should return IRQ_HANDLED. I think that if none of the enabled interrupts happened and got handled, we should return IRQ_NONE -- that way the kernel IRQ code recoginizes a spurious interrupt and masks it off pretty quickly... Fixes: 7729c7a2 ("mmc: tmio: Provide separate interrupt handlers") Signed-off-by:
Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Reviewed-by:
-
Jonathan Neuschäfer authored
commit c9bd505d upstream. When using the mmc_spi driver with a card-detect pin, I noticed that the card was not detected immediately after probe, but only after it was unplugged and plugged back in (and the CD IRQ fired). The call tree looks something like this: mmc_spi_probe mmc_add_host mmc_start_host _mmc_detect_change mmc_schedule_delayed_work(&host->detect, 0) mmc_rescan host->bus_ops->detect(host) mmc_detect _mmc_detect_card_removed host->ops->get_cd(host) mmc_gpio_get_cd -> -ENOSYS (ctx->cd_gpio not set) mmc_gpiod_request_cd ctx->cd_gpio = desc To fix this issue, call mmc_detect_change after the card-detect GPIO/IRQ is registered. Signed-off-by:
Jonathan Neuschäfer <j.neuschaefer@gmx.net> Reviewed-by:
Linus Walleij <linus.walleij@linaro.org> Cc: stable@vger.kernel.org Signed-off-by:
-
Ben Gardon authored
[ Upstream commit 94a980c3 ] Fix a call to userspace_mem_region_find to conform to its spec of taking an inclusive, inclusive range. It was previously being called with an inclusive, exclusive range. Also remove a redundant region bounds check in vm_userspace_mem_region_add. Region overlap checking is already performed by the call to userspace_mem_region_find. Tested: Compiled tools/testing/selftests/kvm with -static Ran all resulting test binaries on an Intel Haswell test machine All tests passed Signed-off-by:
Ben Gardon <bgardon@google.com> Reviewed-by:
Jim Mattson <jmattson@google.com> Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com> Signed-off-by:
Sasha Levin <sashal@kernel.org>
-
Vitaly Kuznetsov authored
[ Upstream commit 619ad846 ] kvm-unit-tests' eventinj "NMI failing on IDT" test results in NMI being delivered to the host (L1) when it's running nested. The problem seems to be: svm_complete_interrupts() raises 'nmi_injected' flag but later we decide to reflect EXIT_NPF to L1. The flag remains pending and we do NMI injection upon entry so it got delivered to L1 instead of L2. It seems that VMX code solves the same issue in prepare_vmcs12(), this was introduced with code refactoring in commit 5f3d5799 ("KVM: nVMX: Rework event injection and recovery"). Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
-
Suravee Suthikulpanit authored
[ Upstream commit bb218fbc ] In case of incomplete IPI with invalid interrupt type, the current SVM driver does not properly emulate the IPI, and fails to boot FreeBSD guests with multiple vcpus when enabling AVIC. Fix this by update APIC ICR high/low registers, which also emulate sending the IPI. Signed-off-by:
Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Signed-off-by:
-
Chaitanya Tata authored
[ Upstream commit 93183bdb ] Recently, DMG frequency bands have been extended till 71GHz, so extend the range check till 20GHz (45-71GHZ), else some channels will be marked as disabled. Signed-off-by:
Chaitanya Tata <Chaitanya.Tata@bluwireless.co.uk> Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Mathieu Malaterre authored
[ Upstream commit 7c53eb5d ] During refactor in commit 9e478066 ("mac80211: fix MU-MIMO follow-MAC mode") a new struct 'action' was declared with packed attribute as: struct { struct ieee80211_hdr_3addr hdr; u8 category; u8 action_code; } __packed action; But since struct 'ieee80211_hdr_3addr' is declared with an aligned keyword as: struct ieee80211_hdr { __le16 frame_control; __le16 duration_id; u8 addr1[ETH_ALEN]; u8 addr2[ETH_ALEN]; u8 addr3[ETH_ALEN]; __le16 seq_ctrl; u8 addr4[ETH_ALEN]; } __packed __aligned(2); Solve the ambiguity of placing aligned structure in a packed one by adding the aligned(2) attribute to struct 'action'. This removes the following warning (W=1): net/mac80211/rx.c:234:2: warning: alignment 1 of 'struct <anonymous>' is less than 2 [-Wpacked-not-aligned] Cc: Johannes Berg <johannes.berg@intel.com> Suggested-by:
Johannes Berg <johannes@sipsolutions.net> Signed-off-by:
Mathieu Malaterre <malat@debian.org> Signed-off-by:
-
Balaji Pothunoori authored
[ Upstream commit 7ed52853 ] Following call trace is observed while adding TDLS peer entry in driver during TDLS setup. Call Trace: [<c1301476>] dump_stack+0x47/0x61 [<c10537d2>] __warn+0xe2/0x100 [<fa22415f>] ? sta_apply_parameters+0x49f/0x550 [mac80211] [<c1053895>] warn_slowpath_null+0x25/0x30 [<fa22415f>] sta_apply_parameters+0x49f/0x550 [mac80211] [<fa20ad42>] ? sta_info_alloc+0x1c2/0x450 [mac80211] [<fa224623>] ieee80211_add_station+0xe3/0x160 [mac80211] [<c1876fe3>] nl80211_new_station+0x273/0x420 [<c170f6d9>] genl_rcv_msg+0x219/0x3c0 [<c170f4c0>] ? genl_rcv+0x30/0x30 [<c170ee7e>] netlink_rcv_skb+0x8e/0xb0 [<c170f4ac>] genl_rcv+0x1c/0x30 [<c170e8aa>] netlink_unicast+0x13a/0x1d0 [<c170ec18>] netlink_sendmsg+0x2d8/0x390 [<c16c5acd>] sock_sendmsg+0x2d/0x40 [<c16c6369>] ___sys_sendmsg+0x1d9/0x1e0 Fixing this by allowing TDLS setup request only when we have completed association. Signed-off-by:
Balaji Pothunoori <bpothuno@codeaurora.org> Signed-off-by:
-
Thomas Falcon authored
[ Upstream commit e95d22c6 ] The IBM virtual ethernet driver's polling function continues to process frames after rescheduling NAPI, resulting in a warning if it exhausted its budget. Do not restart polling after calling napi_reschedule. Instead let frames be processed in the following instance. Signed-off-by:
Thomas Falcon <tlfalcon@linux.ibm.com> Signed-off-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
-
Maciej Żenczykowski authored
[ Upstream commit 3b707c30 ] __bpf_redirect() and act_mirred checks this boolean to determine whether to prefix an ethernet header. Signed-off-by:
Maciej Żenczykowski <maze@google.com> Acked-by:
Daniel Borkmann <daniel@iogearbox.net> Signed-off-by:
-
Zhang Run authored
[ Upstream commit 6eea3527 ] The ax88772_bind() should return error code immediately when the PHY was not reset properly through ax88772a_hw_reset(). Otherwise, The asix_get_phyid() will block when get the PHY Identifier from the PHYSID1 MII registers through asix_mdio_read() due to the PHY isn't ready. Furthermore, it will produce a lot of error message cause system crash.As follows: asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to write reg index 0x0000: -71 asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to send software reset: ffffffb9 asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to write reg index 0x0000: -71 asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to enable software MII access asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to read reg index 0x0000: -71 asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to write reg index 0x0000: -71 asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to enable software MII access asix 1-1:1.0 (unnamed net_device) (uninitialized): Failed to read reg index 0x0000: -71 ... Signed-off-by:
Zhang Run <zhang.run@zte.com.cn> Reviewed-by:
Yang Wei <yang.wei9@zte.com.cn> Tested-by:
Marcel Ziswiler <marcel.ziswiler@toradex.com> Signed-off-by:
-
Douglas Anderson authored
[ Upstream commit a3c5e2cd ] The bindings for Qualcomm opp levels changed after being Acked but before landing. Thus the code in the GPU driver that was relying on the old bindings is now broken. Let's change the code to match the new bindings by adjusting the old string 'qcom,level' to the new string 'opp-level'. See the patch ("dt-bindings: opp: Introduce opp-level bindings"). NOTE: we will do additional cleanup to totally remove the string from the code and use the new dev_pm_opp_get_level() but we'll do it in a future patch. This will facilitate getting the important code fix in sooner without having to deal with cross-maintainer dependencies. This patch needs to land before the patch ("arm64: dts: sdm845: Add gpu and gmu device nodes") since if a tree contains the device tree patch but not this one you'll get a crash at bootup. Fixes: 4b565ca5 ("drm/msm: Add A6XX device support") Signed-off-by:
Douglas Anderson <dianders@chromium.org> Reviewed-by:
Jordan Crouse <jcrouse@codeaurora.org> Signed-off-by:
Rob Clark <robdclark@gmail.com> Signed-off-by:
-
Hannes Reinecke authored
[ Upstream commit 78a61cd4 ] Bit 6 in the ANACAP field is used to indicate that the ANA group ID doesn't change while the namespace is attached to the controller. There is an optimisation in the code to only allocate space for the ANA group header, as the namespace list won't change and hence would not need to be refreshed. However, this optimisation was never carried over to the actual workflow, which always assumes that the buffer is large enough to hold the ANA header _and_ the namespace list. So drop this optimisation and always allocate enough space. Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Hannes Reinecke <hare@suse.com> Signed-off-by:
Sagi Grimberg <sagi@grimberg.me> Signed-off-by:
Jens Axboe <axboe@kernel.dk> Signed-off-by:
-
Sagi Grimberg authored
[ Upstream commit 4c174e63 ] Currently, we have several problems with the timeout handler: 1. If we timeout on the controller establishment flow, we will hang because we don't execute the error recovery (and we shouldn't because the create_ctrl flow needs to fail and cleanup on its own) 2. We might also hang if we get a disconnet on a queue while the controller is already deleting. This racy flow can cause the controller disable/shutdown admin command to hang. We cannot complete a timed out request from the timeout handler without mutual exclusion from the teardown flow (e.g. nvme_rdma_error_recovery_work). So we serialize it in the timeout handler and teardown io and admin queues to guarantee that no one races with us from completing the request. Reported-by:
Jaesoo Lee <jalee@purestorage.com> Reviewed-by:
-
Haiyang Zhang authored
[ Upstream commit 17d91256 ] Changing mtu, channels, or buffer sizes ops call to netvsc_attach(), rndis_set_subchannel(), which always reset the hash key to default value. That will override hash key changed previously. This patch fixes the problem by save the hash key, then restore it when we re- add the netvsc device. Fixes: ff4a4419 ("netvsc: allow get/set of RSS indirection table") Signed-off-by:
Haiyang Zhang <haiyangz@microsoft.com> Reviewed-by:
Michael Kelley <mikelley@microsoft.com> [sl: fix up subject line] Signed-off-by:
-
Haiyang Zhang authored
[ Upstream commit 7c9f335a ] These assignments occur in multiple places. The patch refactor them to a function for simplicity. It also puts the struct to heap area for future expension. Signed-off-by:
-
Haiyang Zhang authored
[ Upstream commit b4a10c75 ] Hyper-V hosts require us to disable RSS before changing RSS key, otherwise the changing request will fail. This patch fixes the coding error. Fixes: ff4a4419 ("netvsc: allow get/set of RSS indirection table") Reported-by:
Wei Hu <weh@microsoft.com> Signed-off-by:
-
Atsushi Nemoto authored
[ Upstream commit 17b42a20 ] The connect_local_phy should return NULL (not negative errno) on error, since its caller expects it. Signed-off-by:
Atsushi Nemoto <atsushi.nemoto@sord.co.jp> Acked-by:
Thor Thayer <thor.thayer@linux.intel.com> Signed-off-by:
-
Varun Prakash authored
[ Upstream commit fe35a40e ] Assign fc_vport to ln->fc_vport before calling csio_fcoe_alloc_vnp() to avoid a NULL pointer dereference in csio_vport_set_state(). ln->fc_vport is dereferenced in csio_vport_set_state(). Signed-off-by:
Varun Prakash <varun@chelsio.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by:
-
Ewan D. Milne authored
[ Upstream commit c41f5988 ] We cannot wait on a completion object in the lpfc_nvme_targetport structure in the _destroy_targetport() code path because the NVMe/fc transport will free that structure immediately after the .targetport_delete() callback. This results in a use-after-free, and a hang if slub_debug=FZPU is enabled. Fix this by putting the completion on the stack. Signed-off-by:
Ewan D. Milne <emilne@redhat.com> Acked-by:
James Smart <james.smart@broadcom.com> Signed-off-by:
-
Ewan D. Milne authored
[ Upstream commit 7961cba6 ] We cannot wait on a completion object in the lpfc_nvme_lport structure in the _destroy_localport() code path because the NVMe/fc transport will free that structure immediately after the .localport_delete() callback. This results in a use-after-free, and a hang if slub_debug=FZPU is enabled. Fix this by putting the completion on the stack. Signed-off-by:
-
Tejun Heo authored
[ Upstream commit 7fc5854f ] sync_inodes_sb() can race against cgwb (cgroup writeback) membership switches and fail to writeback some inodes. For example, if an inode switches to another wb while sync_inodes_sb() is in progress, the new wb might not be visible to bdi_split_work_to_wbs() at all or the inode might jump from a wb which hasn't issued writebacks yet to one which already has. This patch adds backing_dev_info->wb_switch_rwsem to synchronize cgwb switch path against sync_inodes_sb() so that sync_inodes_sb() is guaranteed to see all the target wbs and inodes can't jump wbs to escape syncing. v2: Fixed misplaced rwsem init. Spotted by Jiufei. Signed-off-by:
Tejun Heo <tj@kernel.org> Reported-by:
Jiufei Xue <xuejiufei@gmail.com> Link: http://lkml.kernel.org/r/dc694ae2-f07f-61e1-7097-7c8411cee12d@gmail.comAcked-by:
Jan Kara <jack@suse.cz> Signed-off-by:
-
Ernesto A. Fernández authored
[ Upstream commit 8b9433eb ] On a DIO_SKIP_HOLES filesystem, the ->get_block() method is currently not allowed to create blocks for an empty inode. This confusion comes from trying to bit shift a negative number, so check the size of the inode first. The problem is most visible for hfsplus, because the fallback to buffered I/O doesn't happen and the write fails with EIO. This is in part the fault of the module, because it gives a wrong return value on ->get_block(); that will be fixed in a separate patch. Reviewed-by:
Jeff Moyer <jmoyer@redhat.com> Reviewed-by:
Ernesto A. Fernández <ernesto.mnd.fernandez@gmail.com> Signed-off-by:
-
Liam Mark authored
[ Upstream commit 31eb79db ] Often userspace doesn't know when the kernel will be calling dma_buf_detach on the buffer. If userpace starts its CPU access at the same time as the sg list is being freed it could end up accessing the sg list after it has been freed. Thread A Thread B - DMA_BUF_IOCTL_SYNC IOCT - ion_dma_buf_begin_cpu_access - list_for_each_entry - ion_dma_buf_detatch - free_duped_table - dma_sync_sg_for_cpu Fix this by getting the ion_buffer lock before freeing the sg table memory. Fixes: 2a55e7b5 ("staging: android: ion: Call dma_map_sg for syncing and mapping") Signed-off-by:
Liam Mark <lmark@codeaurora.org> Acked-by:
Laura Abbott <labbott@redhat.com> Acked-by:
Andrew F. Davis <afd@ti.com> Signed-off-by:
-
Priit Laes authored
[ Upstream commit 5e1bc251 ] Although TMDS clock is required for HDMI to properly function, nobody called clk_prepare_enable(). This fixes reference counting issues and makes sure clock is running when it needs to be running. Due to TDMS clock being parent clock for DDC clock, TDMS clock was turned on/off for each EDID probe, causing spurious failures for certain HDMI/DVI screens. Fixes: 9c568101 ("drm/sun4i: Add HDMI support") Signed-off-by:
Priit Laes <priit.laes@paf.com> [Maxime: Moved the TMDS clock enable earlier] Signed-off-by:
Maxime Ripard <maxime.ripard@bootlin.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190122073232.7240-1-plaes@plaes.orgSigned-off-by:
-
Tomonori Sakita authored
[ Upstream commit 815d835b ] Using over-sampling ratio, lpuart can accept baud rate upto uartclk / 4. Signed-off-by:
Tomonori Sakita <tomonori.sakita@sord.co.jp> Signed-off-by:
-
Matthias Kaehlcke authored
[ Upstream commit e8a6ca80 ] The geni set/get_mctrl() functions currently do nothing unless hardware flow control is enabled. Remove this arbitrary limitation. Suggested-by:
Johan Hovold <johan@kernel.org> Fixes: 8a8a66a1 ("tty: serial: qcom_geni_serial: Add support for flow control") Signed-off-by:
Matthias Kaehlcke <mka@chromium.org> Reviewed-by:
-
Kenneth Feng authored
[ Upstream commit 6d87dc97 ] gfxclk for OD setting is limited to 1980M for non-acg ASICs of Vega10 Signed-off-by:
Kenneth Feng <kenneth.feng@amd.com> Reviewed-by:
Evan Quan <evan.quan@amd.com> Signed-off-by:
Alex Deucher <alexander.deucher@amd.com> Signed-off-by:
-
Xie Yongji authored
[ Upstream commit e158488b ] Because wake_q_add() can imply an immediate wakeup (cmpxchg failure case), we must not rely on the wakeup being delayed. However, commit: e3851390 ("locking/rwsem: Rework zeroing reader waiter->task") relies on exactly that behaviour in that the wakeup must not happen until after we clear waiter->task. [ peterz: Added changelog. ] Signed-off-by:
Xie Yongji <xieyongji@baidu.com> Signed-off-by:
Zhang Yu <zhangyu31@baidu.com> Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: e3851390 ("locking/rwsem: Rework zeroing reader waiter->task") Link: https://lkml.kernel.org/r/1543495830-2644-1-git-send-email-xieyongji@baidu.comSigned-off-by:
Ingo Molnar <mingo@kernel.org> Signed-off-by:
-
Peter Zijlstra authored
[ Upstream commit b061c38b ] We must not rely on wake_q_add() to delay the wakeup; in particular commit: 1d0dcb3a ("futex: Implement lockless wakeups") moved wake_q_add() before smp_store_release(&q->lock_ptr, NULL), which could result in futex_wait() waking before observing ->lock_ptr == NULL and going back to sleep again. Signed-off-by:
-
Peter Zijlstra authored
[ Upstream commit 4c4e3731 ] Notable cmpxchg() does not provide ordering when it fails, however wake_q_add() requires ordering in this specific case too. Without this it would be possible for the concurrent wakeup to not observe our prior state. Andrea Parri provided: C wake_up_q-wake_q_add { int next = 0; int y = 0; } P0(int *next, int *y) { int r0; /* in wake_up_q() */ WRITE_ONCE(*next, 1); /* node->next = NULL */ smp_mb(); /* implied by wake_up_process() */ r0 = READ_ONCE(*y); } P1(int *next, int *y) { int r1; /* in wake_q_add() */ WRITE_ONCE(*y, 1); /* wake_cond = true */ smp_mb__before_atomic(); r1 = cmpxchg_relaxed(next, 1, 2); } exists (0:r0=0 /\ 1:r1=0) This "exists" clause cannot be satisfied according to the LKMM: Test wake_up_q-wake_q_add Allowed States 3 0:r0=0; 1:r1=1; 0:r0=1; 1:r1=0; 0:r0=1; 1:r1=1; No Witnesses Positive: 0 Negative: 3 Condition exists (0:r0=0 /\ 1:r1=0) Observation wake_up_q-wake_q_add Never 0 3 Reported-by:
Yongji Xie <elohimes@gmail.com> Signed-off-by:
-
Prateek Sood authored
[ Upstream commit 6dc080ee ] For some peculiar reason rcuwait_wake_up() has the right barrier in the comment, but not in the code. This mistake has been observed to cause a deadlock in the following situation: P1 P2 percpu_up_read() percpu_down_write() rcu_sync_is_idle() // false rcu_sync_enter() ... __percpu_up_read() [S] ,- __this_cpu_dec(*sem->read_count) | smp_rmb(); [L] | task = rcu_dereference(w->task) // NULL | | [S] w->task = current | smp_mb(); | [L] readers_active_check() // fail `-> <store happens here> Where the smp_rmb() (obviously) fails to constrain the store. [ peterz: Added changelog. ] Signed-off-by:
Prateek Sood <prsood@codeaurora.org> Signed-off-by:
Andrea Parri <andrea.parri@amarulasolutions.com> Acked-by:
Davidlohr Bueso <dbueso@suse.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: 8f95c90c ("sched/wait, RCU: Introduce rcuwait machinery") Link: https://lkml.kernel.org/r/1543590656-7157-1-git-send-email-prsood@codeaurora.orgSigned-off-by:
-
Bob Copeland authored
[ Upstream commit a0dc0203 ] In ieee80211_rx_h_mesh_fwding, we increment the 'dropped_frames_ttl' counter when we decrement the ttl to zero. For unicast frames destined for other hosts, we stop processing the frame at that point. For multicast frames, we do not rebroadcast it in this case, but we do pass the frame up the stack to process it on this STA. That doesn't match the usual definition of "dropped," so don't count those as such. With this change, something like `ping6 -i0.2 ff02::1%mesh0` from a peer in a ttl=1 network no longer increments the counter rapidly. Signed-off-by:
Bob Copeland <bobcopeland@fb.com> Signed-off-by:
-
Nathan Chancellor authored
[ Upstream commit 97715058 ] When CONFIG_NO_AUTO_INLINE was present in linux-next (which added '-fno-inline-functions' to KBUILD_CFLAGS), an allyesconfig build with Clang failed at the modpost stage: ERROR: "is_broadcast_mac_addr" [drivers/staging/rtl8723bs/r8723bs.ko] undefined! ERROR: "is_zero_mac_addr" [drivers/staging/rtl8723bs/r8723bs.ko] undefined! ERROR: "is_multicast_mac_addr" [drivers/staging/rtl8723bs/r8723bs.ko] undefined! These functions were marked as extern inline, meaning that if inlining doesn't happen, the function will be undefined, as it is above. This happens to work with GCC because the '-fno-inline-functions' option respects the __inline attribute so all instances of these functions are inlined as expected and the definition doesn't actually matter. However, with Clang and '-fno-inline-functions', a function has to be marked with the __always_inline attribute to be considered for inlining, which none of these functions are. Clang tries to find the symbol definition elsewhere as it was told and fails, which trickles down to modpost. To make sure that this code compiles regardless of compiler and make the intention of the code clearer, use 'static' to ensure these functions are always defined, regardless of inlining. Additionally, silence a checkpatch warning by switching from '__inline' to 'inline'. Signed-off-by:
Nathan Chancellor <natechancellor@gmail.com> Signed-off-by:
-
Aaron Hill authored
[ Upstream commit 129699bb ] Changes since V1: * Use dev_info instead of printk * Use dev_warn instead of BUG_ON Previously, sysfs_create_group was called before all initialization had fully run - specifically, before pci_set_drvdata was called. Since the sysctl group is visible to userspace as soon as sysfs_create_group returns, a small window of time existed during which a process could read from an uninitialized/partially-initialized device. This commit moves the creation of the sysctl group to after all initialized is completed. This ensures that it's impossible for userspace to read from a sysctl file before initialization has fully completed. To catch any future regressions, I've added a check to ensure that proc_thermal_emum_mode is never PROC_THERMAL_NONE when a process tries to read from a sysctl file. Previously, the aforementioned race condition could result in the 'else' branch running while PROC_THERMAL_NONE was set, leading to a null pointer deference. Signed-off-by:
Aaron Hill <aa1ronham@gmail.com> Signed-off-by:
Zhang Rui <rui.zhang@intel.com> Signed-off-by:
-
