January 2018 - Linux-stable-mirror

[Linux-stable-mirror] Patch "x86/acpi: Handle SCI interrupts above legacy space gracefully" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled x86/acpi: Handle SCI interrupts above legacy space gracefully to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: x86-acpi-handle-sci-interrupts-above-legacy-space-gracefully.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 252714155f04c5d16989cb3aadb85fd1b5772f99 Mon Sep 17 00:00:00 2001 From: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> Date: Thu, 16 Nov 2017 21:43:44 +0530 Subject: x86/acpi: Handle SCI interrupts above legacy space gracefully From: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> commit 252714155f04c5d16989cb3aadb85fd1b5772f99 upstream. Platforms which support only IOAPIC mode, pass the SCI information above the legacy space (0-15) via the FADT mechanism and not via MADT. In such cases mp_override_legacy_irq() which is invoked from acpi_sci_ioapic_setup() to register SCI interrupts fails for interrupts greater equal 16, since it is meant to handle only the legacy space and emits error "Invalid bus_irq %u for legacy override". Add a new function to handle SCI interrupts >= 16 and invoke it conditionally in acpi_sci_ioapic_setup(). The code duplication due to this new function will be cleaned up in a separate patch. Co-developed-by: Sunil V L <sunil.vl(a)hpe.com> Signed-off-by: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> Signed-off-by: Sunil V L <sunil.vl(a)hpe.com> Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de> Tested-by: Abdul Lateef Attar <abdul-lateef.attar(a)hpe.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com> Cc: linux-pm(a)vger.kernel.org Cc: kkamagui(a)gmail.com Cc: linux-acpi(a)vger.kernel.org Link: https://lkml.kernel.org/r/1510848825-21965-2-git-send-email-vikas.cha.sajja… Cc: Jean Delvare <jdelvare(a)suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- arch/x86/kernel/acpi/boot.c | 34 +++++++++++++++++++++++++++++++++- 1 file changed, 33 insertions(+), 1 deletion(-) --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -429,6 +429,34 @@ static int mp_config_acpi_gsi(struct dev return 0; } +static int __init mp_register_ioapic_irq(u8 bus_irq, u8 polarity, + u8 trigger, u32 gsi) +{ + struct mpc_intsrc mp_irq; + int ioapic, pin; + + /* Convert 'gsi' to 'ioapic.pin'(INTIN#) */ + ioapic = mp_find_ioapic(gsi); + if (ioapic < 0) { + pr_warn("Failed to find ioapic for gsi : %u\n", gsi); + return ioapic; + } + + pin = mp_find_ioapic_pin(ioapic, gsi); + + mp_irq.type = MP_INTSRC; + mp_irq.irqtype = mp_INT; + mp_irq.irqflag = (trigger << 2) | polarity; + mp_irq.srcbus = MP_ISA_BUS; + mp_irq.srcbusirq = bus_irq; + mp_irq.dstapic = mpc_ioapic_id(ioapic); + mp_irq.dstirq = pin; + + mp_save_irq(&mp_irq); + + return 0; +} + static int __init acpi_parse_ioapic(struct acpi_subtable_header * header, const unsigned long end) { @@ -473,7 +501,11 @@ static void __init acpi_sci_ioapic_setup if (acpi_sci_flags & ACPI_MADT_POLARITY_MASK) polarity = acpi_sci_flags & ACPI_MADT_POLARITY_MASK; - mp_override_legacy_irq(bus_irq, polarity, trigger, gsi); + if (bus_irq < NR_IRQS_LEGACY) + mp_override_legacy_irq(bus_irq, polarity, trigger, gsi); + else + mp_register_ioapic_irq(bus_irq, polarity, trigger, gsi); + acpi_penalize_sci_irq(bus_irq, trigger, polarity); /* Patches currently in stable-queue which might be from vikas.cha.sajjan(a)hpe.com are queue-4.14/x86-acpi-handle-sci-interrupts-above-legacy-space-gracefully.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] Patch "x86/acpi: Handle SCI interrupts above legacy space gracefully" has been added to the 3.18-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled x86/acpi: Handle SCI interrupts above legacy space gracefully to the 3.18-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: x86-acpi-handle-sci-interrupts-above-legacy-space-gracefully.patch and it can be found in the queue-3.18 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 252714155f04c5d16989cb3aadb85fd1b5772f99 Mon Sep 17 00:00:00 2001 From: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> Date: Thu, 16 Nov 2017 21:43:44 +0530 Subject: x86/acpi: Handle SCI interrupts above legacy space gracefully From: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> commit 252714155f04c5d16989cb3aadb85fd1b5772f99 upstream. Platforms which support only IOAPIC mode, pass the SCI information above the legacy space (0-15) via the FADT mechanism and not via MADT. In such cases mp_override_legacy_irq() which is invoked from acpi_sci_ioapic_setup() to register SCI interrupts fails for interrupts greater equal 16, since it is meant to handle only the legacy space and emits error "Invalid bus_irq %u for legacy override". Add a new function to handle SCI interrupts >= 16 and invoke it conditionally in acpi_sci_ioapic_setup(). The code duplication due to this new function will be cleaned up in a separate patch. Co-developed-by: Sunil V L <sunil.vl(a)hpe.com> Signed-off-by: Vikas C Sajjan <vikas.cha.sajjan(a)hpe.com> Signed-off-by: Sunil V L <sunil.vl(a)hpe.com> Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de> Tested-by: Abdul Lateef Attar <abdul-lateef.attar(a)hpe.com> Acked-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com> Cc: linux-pm(a)vger.kernel.org Cc: kkamagui(a)gmail.com Cc: linux-acpi(a)vger.kernel.org Link: https://lkml.kernel.org/r/1510848825-21965-2-git-send-email-vikas.cha.sajja… Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- arch/x86/kernel/acpi/boot.c | 33 ++++++++++++++++++++++++++++++++- 1 file changed, 32 insertions(+), 1 deletion(-) --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -445,6 +445,34 @@ static struct irq_domain_ops acpi_irqdom .unmap = mp_irqdomain_unmap, }; +static int __init mp_register_ioapic_irq(u8 bus_irq, u8 polarity, + u8 trigger, u32 gsi) +{ + struct mpc_intsrc mp_irq; + int ioapic, pin; + + /* Convert 'gsi' to 'ioapic.pin'(INTIN#) */ + ioapic = mp_find_ioapic(gsi); + if (ioapic < 0) { + pr_warn("Failed to find ioapic for gsi : %u\n", gsi); + return ioapic; + } + + pin = mp_find_ioapic_pin(ioapic, gsi); + + mp_irq.type = MP_INTSRC; + mp_irq.irqtype = mp_INT; + mp_irq.irqflag = (trigger << 2) | polarity; + mp_irq.srcbus = MP_ISA_BUS; + mp_irq.srcbusirq = bus_irq; + mp_irq.dstapic = mpc_ioapic_id(ioapic); + mp_irq.dstirq = pin; + + mp_save_irq(&mp_irq); + + return 0; +} + static int __init acpi_parse_ioapic(struct acpi_subtable_header * header, const unsigned long end) { @@ -489,7 +517,10 @@ static void __init acpi_sci_ioapic_setup if (acpi_sci_flags & ACPI_MADT_POLARITY_MASK) polarity = acpi_sci_flags & ACPI_MADT_POLARITY_MASK; - mp_override_legacy_irq(bus_irq, polarity, trigger, gsi); + if (bus_irq < NR_IRQS_LEGACY) + mp_override_legacy_irq(bus_irq, polarity, trigger, gsi); + else + mp_register_ioapic_irq(bus_irq, polarity, trigger, gsi); /* * stash over-ride to indicate we've been here Patches currently in stable-queue which might be from vikas.cha.sajjan(a)hpe.com are queue-3.18/x86-acpi-handle-sci-interrupts-above-legacy-space-gracefully.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] v4.9.76 build: 0 failures 0 warnings (v4.9.76)

by Build bot for Mark Brown

Tree/Branch: v4.9.76 Git describe: v4.9.76 Commit: 7bbc6ca488 Linux 4.9.76 Build Time: 92 min 21 sec Passed: 10 / 10 (100.00 %) Failed: 0 / 10 ( 0.00 %) Errors: 0 Warnings: 0 Section Mismatches: 0 ------------------------------------------------------------------------------- defconfigs with issues (other than build errors): ------------------------------------------------------------------------------- =============================================================================== Detailed per-defconfig build reports below: ------------------------------------------------------------------------------- Passed with no errors, warnings or mismatches: arm64-allnoconfig arm64-allmodconfig arm-multi_v5_defconfig arm-multi_v7_defconfig x86_64-defconfig arm-allmodconfig arm-allnoconfig x86_64-allnoconfig arm-multi_v4t_defconfig arm64-defconfig close failed in file object destructor: sys.excepthook is missing lost sys.stderr

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] [PATCH] kselftest: fix OOM in memory compaction test

by Arnd Bergmann

Running the compaction_test sometimes results in out-of-memory failures. When I debugged this, it turned out that the code to reset the number of hugepages to the initial value is simply broken since we write into an open sysctl file descriptor multiple times without seeking back to the start. Adding the lseek here fixes the problem. Cc: stable(a)vger.kernel.org Reported-by: Naresh Kamboju <naresh.kamboju(a)linaro.org> Link: https://bugs.linaro.org/show_bug.cgi?id=3145 Signed-off-by: Arnd Bergmann <arnd(a)arndb.de> --- tools/testing/selftests/vm/compaction_test.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/testing/selftests/vm/compaction_test.c b/tools/testing/selftests/vm/compaction_test.c index a65b016d4c13..1097f04e4d80 100644 --- a/tools/testing/selftests/vm/compaction_test.c +++ b/tools/testing/selftests/vm/compaction_test.c @@ -137,6 +137,8 @@ int check_compaction(unsigned long mem_free, unsigned int hugepage_size) printf("No of huge pages allocated = %d\n", (atoi(nr_hugepages))); + lseek(fd, 0, SEEK_SET); + if (write(fd, initial_nr_hugepages, strlen(initial_nr_hugepages)) != strlen(initial_nr_hugepages)) { perror("Failed to write value to /proc/sys/vm/nr_hugepages\n"); -- 2.9.0

6 years, 11 months

2
2
0 0

[Linux-stable-mirror] Patch "iw_cxgb4: reflect the original WR opcode in drain cqes" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled iw_cxgb4: reflect the original WR opcode in drain cqes to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 96a236ed286776554fbd227c6d2876fd3b5dc65d Mon Sep 17 00:00:00 2001 From: Steve Wise <swise(a)opengridcomputing.com> Date: Tue, 19 Dec 2017 10:29:25 -0800 Subject: iw_cxgb4: reflect the original WR opcode in drain cqes From: Steve Wise <swise(a)opengridcomputing.com> commit 96a236ed286776554fbd227c6d2876fd3b5dc65d upstream. The flush/drain logic was not retaining the original wr opcode in its completion. This can cause problems if the application uses the completion opcode to make decisions. Use bit 10 of the CQE header word to indicate the CQE is a special drain completion, and save the original WR opcode in the cqe header opcode field. Fixes: 4fe7c2962e11 ("iw_cxgb4: refactor sq/rq drain logic") Signed-off-by: Steve Wise <swise(a)opengridcomputing.com> Signed-off-by: Jason Gunthorpe <jgg(a)mellanox.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/infiniband/hw/cxgb4/cq.c | 7 +---- drivers/infiniband/hw/cxgb4/iw_cxgb4.h | 2 - drivers/infiniband/hw/cxgb4/qp.c | 46 ++++++++++++++++++++++++++++++--- drivers/infiniband/hw/cxgb4/t4.h | 6 ++++ 4 files changed, 50 insertions(+), 11 deletions(-) --- a/drivers/infiniband/hw/cxgb4/cq.c +++ b/drivers/infiniband/hw/cxgb4/cq.c @@ -410,7 +410,7 @@ next_cqe: static int cqe_completes_wr(struct t4_cqe *cqe, struct t4_wq *wq) { - if (CQE_OPCODE(cqe) == C4IW_DRAIN_OPCODE) { + if (DRAIN_CQE(cqe)) { WARN_ONCE(1, "Unexpected DRAIN CQE qp id %u!\n", wq->sq.qid); return 0; } @@ -509,7 +509,7 @@ static int poll_cq(struct t4_wq *wq, str /* * Special cqe for drain WR completions... */ - if (CQE_OPCODE(hw_cqe) == C4IW_DRAIN_OPCODE) { + if (DRAIN_CQE(hw_cqe)) { *cookie = CQE_DRAIN_COOKIE(hw_cqe); *cqe = *hw_cqe; goto skip_cqe; @@ -766,9 +766,6 @@ static int c4iw_poll_cq_one(struct c4iw_ c4iw_invalidate_mr(qhp->rhp, CQE_WRID_FR_STAG(&cqe)); break; - case C4IW_DRAIN_OPCODE: - wc->opcode = IB_WC_SEND; - break; default: pr_err("Unexpected opcode %d in the CQE received for QPID=0x%0x\n", CQE_OPCODE(&cqe), CQE_QPID(&cqe)); --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -631,8 +631,6 @@ static inline int to_ib_qp_state(int c4i return IB_QPS_ERR; } -#define C4IW_DRAIN_OPCODE FW_RI_SGE_EC_CR_RETURN - static inline u32 c4iw_ib_to_tpt_access(int a) { return (a & IB_ACCESS_REMOTE_WRITE ? FW_RI_MEM_ACCESS_REM_WRITE : 0) | --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -794,21 +794,57 @@ static int ring_kernel_rq_db(struct c4iw return 0; } -static void complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr) +static int ib_to_fw_opcode(int ib_opcode) +{ + int opcode; + + switch (ib_opcode) { + case IB_WR_SEND_WITH_INV: + opcode = FW_RI_SEND_WITH_INV; + break; + case IB_WR_SEND: + opcode = FW_RI_SEND; + break; + case IB_WR_RDMA_WRITE: + opcode = FW_RI_RDMA_WRITE; + break; + case IB_WR_RDMA_READ: + case IB_WR_RDMA_READ_WITH_INV: + opcode = FW_RI_READ_REQ; + break; + case IB_WR_REG_MR: + opcode = FW_RI_FAST_REGISTER; + break; + case IB_WR_LOCAL_INV: + opcode = FW_RI_LOCAL_INV; + break; + default: + opcode = -EINVAL; + } + return opcode; +} + +static int complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr) { struct t4_cqe cqe = {}; struct c4iw_cq *schp; unsigned long flag; struct t4_cq *cq; + int opcode; schp = to_c4iw_cq(qhp->ibqp.send_cq); cq = &schp->cq; + opcode = ib_to_fw_opcode(wr->opcode); + if (opcode < 0) + return opcode; + cqe.u.drain_cookie = wr->wr_id; cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) | - CQE_OPCODE_V(C4IW_DRAIN_OPCODE) | + CQE_OPCODE_V(opcode) | CQE_TYPE_V(1) | CQE_SWCQE_V(1) | + CQE_DRAIN_V(1) | CQE_QPID_V(qhp->wq.sq.qid)); spin_lock_irqsave(&schp->lock, flag); @@ -823,6 +859,7 @@ static void complete_sq_drain_wr(struct schp->ibcq.cq_context); spin_unlock_irqrestore(&schp->comp_handler_lock, flag); } + return 0; } static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr) @@ -837,9 +874,10 @@ static void complete_rq_drain_wr(struct cqe.u.drain_cookie = wr->wr_id; cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) | - CQE_OPCODE_V(C4IW_DRAIN_OPCODE) | + CQE_OPCODE_V(FW_RI_SEND) | CQE_TYPE_V(0) | CQE_SWCQE_V(1) | + CQE_DRAIN_V(1) | CQE_QPID_V(qhp->wq.sq.qid)); spin_lock_irqsave(&rchp->lock, flag); @@ -879,7 +917,7 @@ int c4iw_post_send(struct ib_qp *ibqp, s */ if (qhp->wq.flushed) { spin_unlock_irqrestore(&qhp->lock, flag); - complete_sq_drain_wr(qhp, wr); + err = complete_sq_drain_wr(qhp, wr); return err; } num_wrs = t4_sq_avail(&qhp->wq); --- a/drivers/infiniband/hw/cxgb4/t4.h +++ b/drivers/infiniband/hw/cxgb4/t4.h @@ -197,6 +197,11 @@ struct t4_cqe { #define CQE_SWCQE_G(x) ((((x) >> CQE_SWCQE_S)) & CQE_SWCQE_M) #define CQE_SWCQE_V(x) ((x)<<CQE_SWCQE_S) +#define CQE_DRAIN_S 10 +#define CQE_DRAIN_M 0x1 +#define CQE_DRAIN_G(x) ((((x) >> CQE_DRAIN_S)) & CQE_DRAIN_M) +#define CQE_DRAIN_V(x) ((x)<<CQE_DRAIN_S) + #define CQE_STATUS_S 5 #define CQE_STATUS_M 0x1F #define CQE_STATUS_G(x) ((((x) >> CQE_STATUS_S)) & CQE_STATUS_M) @@ -213,6 +218,7 @@ struct t4_cqe { #define CQE_OPCODE_V(x) ((x)<<CQE_OPCODE_S) #define SW_CQE(x) (CQE_SWCQE_G(be32_to_cpu((x)->header))) +#define DRAIN_CQE(x) (CQE_DRAIN_G(be32_to_cpu((x)->header))) #define CQE_QPID(x) (CQE_QPID_G(be32_to_cpu((x)->header))) #define CQE_TYPE(x) (CQE_TYPE_G(be32_to_cpu((x)->header))) #define SQ_TYPE(x) (CQE_TYPE((x))) Patches currently in stable-queue which might be from swise(a)opengridcomputing.com are queue-4.14/iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch queue-4.14/iw_cxgb4-atomically-flush-the-qp.patch queue-4.14/iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch queue-4.14/iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch queue-4.14/iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] Patch "iw_cxgb4: when flushing, complete all wrs in a chain" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled iw_cxgb4: when flushing, complete all wrs in a chain to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From d14587334580bc94d3ee11e8320e0c157f91ae8f Mon Sep 17 00:00:00 2001 From: Steve Wise <swise(a)opengridcomputing.com> Date: Tue, 19 Dec 2017 14:02:10 -0800 Subject: iw_cxgb4: when flushing, complete all wrs in a chain From: Steve Wise <swise(a)opengridcomputing.com> commit d14587334580bc94d3ee11e8320e0c157f91ae8f upstream. If a wr chain was posted and needed to be flushed, only the first wr in the chain was completed with FLUSHED status. The rest were never completed. This caused isert to hang on shutdown due to the missing completions which left iscsi IO commands referenced, stalling the shutdown. Fixes: 4fe7c2962e11 ("iw_cxgb4: refactor sq/rq drain logic") Signed-off-by: Steve Wise <swise(a)opengridcomputing.com> Signed-off-by: Jason Gunthorpe <jgg(a)mellanox.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/infiniband/hw/cxgb4/qp.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -862,6 +862,22 @@ static int complete_sq_drain_wr(struct c return 0; } +static int complete_sq_drain_wrs(struct c4iw_qp *qhp, struct ib_send_wr *wr, + struct ib_send_wr **bad_wr) +{ + int ret = 0; + + while (wr) { + ret = complete_sq_drain_wr(qhp, wr); + if (ret) { + *bad_wr = wr; + break; + } + wr = wr->next; + } + return ret; +} + static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr) { struct t4_cqe cqe = {}; @@ -894,6 +910,14 @@ static void complete_rq_drain_wr(struct } } +static void complete_rq_drain_wrs(struct c4iw_qp *qhp, struct ib_recv_wr *wr) +{ + while (wr) { + complete_rq_drain_wr(qhp, wr); + wr = wr->next; + } +} + int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, struct ib_send_wr **bad_wr) { @@ -917,7 +941,7 @@ int c4iw_post_send(struct ib_qp *ibqp, s */ if (qhp->wq.flushed) { spin_unlock_irqrestore(&qhp->lock, flag); - err = complete_sq_drain_wr(qhp, wr); + err = complete_sq_drain_wrs(qhp, wr, bad_wr); return err; } num_wrs = t4_sq_avail(&qhp->wq); @@ -1066,7 +1090,7 @@ int c4iw_post_receive(struct ib_qp *ibqp */ if (qhp->wq.flushed) { spin_unlock_irqrestore(&qhp->lock, flag); - complete_rq_drain_wr(qhp, wr); + complete_rq_drain_wrs(qhp, wr); return err; } num_wrs = t4_rq_avail(&qhp->wq); Patches currently in stable-queue which might be from swise(a)opengridcomputing.com are queue-4.14/iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch queue-4.14/iw_cxgb4-atomically-flush-the-qp.patch queue-4.14/iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch queue-4.14/iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch queue-4.14/iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] Patch "iw_cxgb4: only clear the ARMED bit if a notification is needed" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled iw_cxgb4: only clear the ARMED bit if a notification is needed to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 335ebf6fa35ca1c59b73f76fad19b249d3550e86 Mon Sep 17 00:00:00 2001 From: Steve Wise <swise(a)opengridcomputing.com> Date: Thu, 30 Nov 2017 09:41:56 -0800 Subject: iw_cxgb4: only clear the ARMED bit if a notification is needed From: Steve Wise <swise(a)opengridcomputing.com> commit 335ebf6fa35ca1c59b73f76fad19b249d3550e86 upstream. In __flush_qp(), the CQ ARMED bit was being cleared regardless of whether any notification is actually needed. This resulted in the iser termination logic getting stuck in ib_drain_sq() because the CQ was not marked ARMED and thus the drain CQE notification wasn't triggered. This new bug was exposed when this commit was merged: commit cbb40fadd31c ("iw_cxgb4: only call the cq comp_handler when the cq is armed") Signed-off-by: Steve Wise <swise(a)opengridcomputing.com> Signed-off-by: Jason Gunthorpe <jgg(a)mellanox.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/infiniband/hw/cxgb4/qp.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -1301,21 +1301,21 @@ static void __flush_qp(struct c4iw_qp *q spin_unlock_irqrestore(&rchp->lock, flag); if (schp == rchp) { - if (t4_clear_cq_armed(&rchp->cq) && - (rq_flushed || sq_flushed)) { + if ((rq_flushed || sq_flushed) && + t4_clear_cq_armed(&rchp->cq)) { spin_lock_irqsave(&rchp->comp_handler_lock, flag); (*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context); spin_unlock_irqrestore(&rchp->comp_handler_lock, flag); } } else { - if (t4_clear_cq_armed(&rchp->cq) && rq_flushed) { + if (rq_flushed && t4_clear_cq_armed(&rchp->cq)) { spin_lock_irqsave(&rchp->comp_handler_lock, flag); (*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context); spin_unlock_irqrestore(&rchp->comp_handler_lock, flag); } - if (t4_clear_cq_armed(&schp->cq) && sq_flushed) { + if (sq_flushed && t4_clear_cq_armed(&schp->cq)) { spin_lock_irqsave(&schp->comp_handler_lock, flag); (*schp->ibcq.comp_handler)(&schp->ibcq, schp->ibcq.cq_context); Patches currently in stable-queue which might be from swise(a)opengridcomputing.com are queue-4.14/iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch queue-4.14/iw_cxgb4-atomically-flush-the-qp.patch queue-4.14/iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch queue-4.14/iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch queue-4.14/iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] Patch "iw_cxgb4: only call the cq comp_handler when the cq is armed" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled iw_cxgb4: only call the cq comp_handler when the cq is armed to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From cbb40fadd31c6bbc59104e58ac95c6ef492d038b Mon Sep 17 00:00:00 2001 From: Steve Wise <swise(a)opengridcomputing.com> Date: Thu, 9 Nov 2017 07:14:43 -0800 Subject: iw_cxgb4: only call the cq comp_handler when the cq is armed From: Steve Wise <swise(a)opengridcomputing.com> commit cbb40fadd31c6bbc59104e58ac95c6ef492d038b upstream. The ULPs completion handler should only be called if the CQ is armed for notification. Signed-off-by: Steve Wise <swise(a)opengridcomputing.com> Signed-off-by: Doug Ledford <dledford(a)redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/infiniband/hw/cxgb4/ev.c | 8 +++++--- drivers/infiniband/hw/cxgb4/qp.c | 20 ++++++++++++-------- 2 files changed, 17 insertions(+), 11 deletions(-) --- a/drivers/infiniband/hw/cxgb4/ev.c +++ b/drivers/infiniband/hw/cxgb4/ev.c @@ -109,9 +109,11 @@ static void post_qp_event(struct c4iw_de if (qhp->ibqp.event_handler) (*qhp->ibqp.event_handler)(&event, qhp->ibqp.qp_context); - spin_lock_irqsave(&chp->comp_handler_lock, flag); - (*chp->ibcq.comp_handler)(&chp->ibcq, chp->ibcq.cq_context); - spin_unlock_irqrestore(&chp->comp_handler_lock, flag); + if (t4_clear_cq_armed(&chp->cq)) { + spin_lock_irqsave(&chp->comp_handler_lock, flag); + (*chp->ibcq.comp_handler)(&chp->ibcq, chp->ibcq.cq_context); + spin_unlock_irqrestore(&chp->comp_handler_lock, flag); + } } void c4iw_ev_dispatch(struct c4iw_dev *dev, struct t4_cqe *err_cqe) --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -817,10 +817,12 @@ static void complete_sq_drain_wr(struct t4_swcq_produce(cq); spin_unlock_irqrestore(&schp->lock, flag); - spin_lock_irqsave(&schp->comp_handler_lock, flag); - (*schp->ibcq.comp_handler)(&schp->ibcq, - schp->ibcq.cq_context); - spin_unlock_irqrestore(&schp->comp_handler_lock, flag); + if (t4_clear_cq_armed(&schp->cq)) { + spin_lock_irqsave(&schp->comp_handler_lock, flag); + (*schp->ibcq.comp_handler)(&schp->ibcq, + schp->ibcq.cq_context); + spin_unlock_irqrestore(&schp->comp_handler_lock, flag); + } } static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr) @@ -846,10 +848,12 @@ static void complete_rq_drain_wr(struct t4_swcq_produce(cq); spin_unlock_irqrestore(&rchp->lock, flag); - spin_lock_irqsave(&rchp->comp_handler_lock, flag); - (*rchp->ibcq.comp_handler)(&rchp->ibcq, - rchp->ibcq.cq_context); - spin_unlock_irqrestore(&rchp->comp_handler_lock, flag); + if (t4_clear_cq_armed(&rchp->cq)) { + spin_lock_irqsave(&rchp->comp_handler_lock, flag); + (*rchp->ibcq.comp_handler)(&rchp->ibcq, + rchp->ibcq.cq_context); + spin_unlock_irqrestore(&rchp->comp_handler_lock, flag); + } } int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, Patches currently in stable-queue which might be from swise(a)opengridcomputing.com are queue-4.14/iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch queue-4.14/iw_cxgb4-atomically-flush-the-qp.patch queue-4.14/iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch queue-4.14/iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch queue-4.14/iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] Patch "iw_cxgb4: atomically flush the qp" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled iw_cxgb4: atomically flush the qp to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: iw_cxgb4-atomically-flush-the-qp.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From bc52e9ca74b9a395897bb640c6671b2cbf716032 Mon Sep 17 00:00:00 2001 From: Steve Wise <swise(a)opengridcomputing.com> Date: Thu, 9 Nov 2017 07:21:26 -0800 Subject: iw_cxgb4: atomically flush the qp From: Steve Wise <swise(a)opengridcomputing.com> commit bc52e9ca74b9a395897bb640c6671b2cbf716032 upstream. __flush_qp() has a race condition where during the flush operation, the qp lock is released allowing another thread to possibly post a WR, which corrupts the queue state, possibly causing crashes. The lock was released to preserve the cq/qp locking hierarchy of cq first, then qp. However releasing the qp lock is not necessary; both RQ and SQ CQ locks can be acquired first, followed by the qp lock, and then the RQ and SQ flushing can be done w/o unlocking. Signed-off-by: Steve Wise <swise(a)opengridcomputing.com> Signed-off-by: Doug Ledford <dledford(a)redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/infiniband/hw/cxgb4/qp.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -1271,31 +1271,34 @@ static void __flush_qp(struct c4iw_qp *q pr_debug("%s qhp %p rchp %p schp %p\n", __func__, qhp, rchp, schp); - /* locking hierarchy: cq lock first, then qp lock. */ + /* locking hierarchy: cqs lock first, then qp lock. */ spin_lock_irqsave(&rchp->lock, flag); + if (schp != rchp) + spin_lock(&schp->lock); spin_lock(&qhp->lock); if (qhp->wq.flushed) { spin_unlock(&qhp->lock); + if (schp != rchp) + spin_unlock(&schp->lock); spin_unlock_irqrestore(&rchp->lock, flag); return; } qhp->wq.flushed = 1; + t4_set_wq_in_error(&qhp->wq); c4iw_flush_hw_cq(rchp); c4iw_count_rcqes(&rchp->cq, &qhp->wq, &count); rq_flushed = c4iw_flush_rq(&qhp->wq, &rchp->cq, count); - spin_unlock(&qhp->lock); - spin_unlock_irqrestore(&rchp->lock, flag); - /* locking hierarchy: cq lock first, then qp lock. */ - spin_lock_irqsave(&schp->lock, flag); - spin_lock(&qhp->lock); if (schp != rchp) c4iw_flush_hw_cq(schp); sq_flushed = c4iw_flush_sq(qhp); + spin_unlock(&qhp->lock); - spin_unlock_irqrestore(&schp->lock, flag); + if (schp != rchp) + spin_unlock(&schp->lock); + spin_unlock_irqrestore(&rchp->lock, flag); if (schp == rchp) { if (t4_clear_cq_armed(&rchp->cq) && @@ -1329,8 +1332,8 @@ static void flush_qp(struct c4iw_qp *qhp rchp = to_c4iw_cq(qhp->ibqp.recv_cq); schp = to_c4iw_cq(qhp->ibqp.send_cq); - t4_set_wq_in_error(&qhp->wq); if (qhp->ibqp.uobject) { + t4_set_wq_in_error(&qhp->wq); t4_set_cq_in_error(&rchp->cq); spin_lock_irqsave(&rchp->comp_handler_lock, flag); (*rchp->ibcq.comp_handler)(&rchp->ibcq, rchp->ibcq.cq_context); Patches currently in stable-queue which might be from swise(a)opengridcomputing.com are queue-4.14/iw_cxgb4-only-clear-the-armed-bit-if-a-notification-is-needed.patch queue-4.14/iw_cxgb4-atomically-flush-the-qp.patch queue-4.14/iw_cxgb4-when-flushing-complete-all-wrs-in-a-chain.patch queue-4.14/iw_cxgb4-reflect-the-original-wr-opcode-in-drain-cqes.patch queue-4.14/iw_cxgb4-only-call-the-cq-comp_handler-when-the-cq-is-armed.patch

6 years, 11 months

1
0
0 0

[Linux-stable-mirror] [RESEND PATCH 0/5] iw_cxgb4 - flush bug fixes

by Steve Wise

I'm resending with a cover letter as I realize I didn't provide enough info. This series fixes critical qp flush bugs in iw_cxgb4. Pleaes apply to 4.14-stable. Thanks, Steve ---- Steve Wise (5): iw_cxgb4: only call the cq comp_handler when the cq is armed iw_cxgb4: atomically flush the qp iw_cxgb4: only clear the ARMED bit if a notification is needed iw_cxgb4: reflect the original WR opcode in drain cqes iw_cxgb4: when flushing, complete all wrs in a chain drivers/infiniband/hw/cxgb4/cq.c | 7 +- drivers/infiniband/hw/cxgb4/ev.c | 8 ++- drivers/infiniband/hw/cxgb4/iw_cxgb4.h | 2 - drivers/infiniband/hw/cxgb4/qp.c | 119 ++++++++++++++++++++++++++------- drivers/infiniband/hw/cxgb4/t4.h | 6 ++ 5 files changed, 107 insertions(+), 35 deletions(-) -- 1.8.3.1

6 years, 11 months

2
6
0 0

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror January 2018