From: Yu Kuai yukuai3@huawei.com

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I65K8D CVE: NA

--------------------------------

Before commit f60df4a0a6ad ("blk-mq: fix kabi broken in struct request"), drivers will got cmd address right after request, however, after this commit, drivers will got cmd address after request_wrapper instead, which is bigger than request and will cause compatibility issues.

Fix the problem by placing request_wrapper behind cmd, so that the cmd address for drivers will stay the same.

Before commit: |request|cmd| After commit: |request|request_wrapper|cmd| With this patch: |request|cmd|request_wrapper|

Performance test: arm64 Kunpeng-920 96 core

1) null_blk setup: modprobe null_blk nr_devices=0 && udevadm settle && cd /sys/kernel/config/nullb && mkdir nullb0 && cd nullb0 && echo 0 > completion_nsec && echo 512 > blocksize && echo 0 > home_node && echo 0 > irqmode && echo 1024 > size && echo 0 > memory_backed && echo 2 > queue_mode && echo 4096 > hw_queue_depth && echo 96 > submit_queues && echo 1 > power

2) fio test script: [global] ioengine=libaio direct=1 numjobs=96 iodepth=32 bs=4k rw=randwrite allow_mounted_write=0 time_based runtime=60 group_reporting=1 ioscheduler=none cpus_allowed_policy=split cpus_allowed=0-95

[test] filename=/dev/nullb0

3) iops test result:

without this patch: 23.9M with this patch: 24.1M

Fixes: f60df4a0a6ad ("blk-mq: fix kabi broken in struct request") Signed-off-by: Yu Kuai yukuai3@huawei.com Reviewed-by: Hou Tao houtao1@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- block/blk-flush.c | 2 +- block/blk-mq.c | 5 +++-- block/blk-mq.h | 13 +++++++++++++ drivers/scsi/scsi_error.c | 2 +- include/linux/blk-mq.h | 13 ++----------- 5 files changed, 20 insertions(+), 15 deletions(-)

diff --git a/block/blk-flush.c b/block/blk-flush.c index 71faf07a626f..767624910270 100644 --- a/block/blk-flush.c +++ b/block/blk-flush.c @@ -470,7 +470,7 @@ struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size, gfp_t flags) { struct blk_flush_queue *fq; - int rq_sz = sizeof(struct request_wrapper); + int rq_sz = sizeof(struct request) + sizeof(struct request_wrapper);

fq = kzalloc_node(sizeof(*fq), flags, node); if (!fq) diff --git a/block/blk-mq.c b/block/blk-mq.c index b9b2d9412b02..4c719469ba8e 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -2601,8 +2601,9 @@ static int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, * rq_size is the size of the request plus driver payload, rounded * to the cacheline size */ - rq_size = round_up(sizeof(struct request_wrapper) + set->cmd_size, - cache_line_size()); + rq_size = round_up(sizeof(struct request) + + sizeof(struct request_wrapper) + set->cmd_size, + cache_line_size()); left = rq_size * depth;

for (i = 0; i < depth; ) { diff --git a/block/blk-mq.h b/block/blk-mq.h index 1c86f7d56e72..6254abe9c112 100644 --- a/block/blk-mq.h +++ b/block/blk-mq.h @@ -37,6 +37,19 @@ struct blk_mq_ctx { struct kobject kobj; } ____cacheline_aligned_in_smp;

+struct request_wrapper { + /* Time that I/O was counted in part_get_stat_info(). */ + u64 stat_time_ns; +}; + +static inline struct request_wrapper *request_to_wrapper(struct request *rq) +{ + unsigned long addr = (unsigned long)rq; + + addr += sizeof(*rq) + rq->q->tag_set->cmd_size; + return (struct request_wrapper *)addr; +} + void blk_mq_exit_queue(struct request_queue *q); int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr); void blk_mq_wake_waiters(struct request_queue *q); diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c index bcbeadb2d0f0..f11f51e2465f 100644 --- a/drivers/scsi/scsi_error.c +++ b/drivers/scsi/scsi_error.c @@ -2359,7 +2359,7 @@ scsi_ioctl_reset(struct scsi_device *dev, int __user *arg) return -EIO;

error = -EIO; - rq = kzalloc(sizeof(struct request_wrapper) + sizeof(struct scsi_cmnd) + + rq = kzalloc(sizeof(struct request) + sizeof(struct scsi_cmnd) + shost->hostt->cmd_size, GFP_KERNEL); if (!rq) goto out_put_autopm_host; diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h index aa0c8ef9a50f..f3d78b939e01 100644 --- a/include/linux/blk-mq.h +++ b/include/linux/blk-mq.h @@ -303,15 +303,6 @@ struct blk_mq_queue_data { KABI_RESERVE(1) };

-struct request_wrapper { - struct request rq; - - /* Time that I/O was counted in part_get_stat_info(). */ - u64 stat_time_ns; -}; - -#define request_to_wrapper(_rq) container_of(_rq, struct request_wrapper, rq) - typedef bool (busy_iter_fn)(struct blk_mq_hw_ctx *, struct request *, void *, bool); typedef bool (busy_tag_iter_fn)(struct request *, void *, bool); @@ -606,7 +597,7 @@ static inline bool blk_should_fake_timeout(struct request_queue *q) */ static inline struct request *blk_mq_rq_from_pdu(void *pdu) { - return pdu - sizeof(struct request_wrapper); + return pdu - sizeof(struct request); }

/** @@ -620,7 +611,7 @@ static inline struct request *blk_mq_rq_from_pdu(void *pdu) */ static inline void *blk_mq_rq_to_pdu(struct request *rq) { - return request_to_wrapper(rq) + 1; + return rq + 1; }

static inline struct blk_mq_hw_ctx *queue_hctx(struct request_queue *q, int id)

From: Junhao He hejunhao3@huawei.com

driver inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5YCYK

--------------------------------------------------------------------------

TRBE device can use the id_tabel to match driver. So if the CoreSight component does not use the device tree or ACPI to report devices (@fwnode is NULL), we need to return the pointer to the platform data @pdata.

Signed-off-by: Junhao He hejunhao3@huawei.com Reviewed-by: Xiongfeng Wang wangxiongfeng2@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- drivers/hwtracing/coresight/coresight-platform.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/hwtracing/coresight/coresight-platform.c b/drivers/hwtracing/coresight/coresight-platform.c index c594f45319fc..82b900076dbd 100644 --- a/drivers/hwtracing/coresight/coresight-platform.c +++ b/drivers/hwtracing/coresight/coresight-platform.c @@ -844,15 +844,15 @@ coresight_get_platform_data(struct device *dev) struct coresight_platform_data *pdata = NULL; struct fwnode_handle *fwnode = dev_fwnode(dev);

- if (IS_ERR_OR_NULL(fwnode)) - goto error; - pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL); if (!pdata) { ret = -ENOMEM; goto error; }

+ if (IS_ERR_OR_NULL(fwnode)) + return pdata; + if (is_of_node(fwnode)) ret = of_get_coresight_platform_data(dev, pdata); else if (is_acpi_device_node(fwnode))

From: Junhao He hejunhao3@huawei.com

driver inclusion category: featurn bugzilla: https://gitee.com/openeuler/kernel/issues/I5YCYK

--------------------------------------------------------------------------

ACPI 6.5 adds additional fields to the MADT GICC structure to describe TRBE PPI's. We pick these out of the cached reference to the madt_gicc structure similarly to the arm SPE code. We then create a platform device referring to the IRQ and let the user/module loader decide whether to load the TRBE driver.

Signed-off-by: Junhao He hejunhao3@huawei.com Reviewed-by: Xiongfeng Wang wangxiongfeng2@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- arch/arm64/configs/openeuler_defconfig | 1 + arch/arm64/include/asm/acpi.h | 3 + arch/arm64/kernel/Makefile | 1 + arch/arm64/kernel/acpi_trbe.c | 81 ++++++++++++++++++++++++++ drivers/hwtracing/coresight/Kconfig | 4 ++ include/linux/coresight.h | 2 + 6 files changed, 92 insertions(+) create mode 100644 arch/arm64/kernel/acpi_trbe.c

diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index b276b14fc5c8..30384eacad4f 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -7299,6 +7299,7 @@ CONFIG_ETM4X_IMPDEF_FEATURE=y # CONFIG_CORESIGHT_CTI is not set CONFIG_CORESIGHT_TRBE=m CONFIG_ULTRASOC_SMB=m +CONFIG_ACPI_TRBE=y # end of arm64 Debugging

# diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h index df0ef7ec5b5b..8f67d367e381 100644 --- a/arch/arm64/include/asm/acpi.h +++ b/arch/arm64/include/asm/acpi.h @@ -42,6 +42,9 @@ #define ACPI_MADT_GICC_SPE (offsetof(struct acpi_madt_generic_interrupt, \ spe_interrupt) + sizeof(u16))

+#define ACPI_MADT_GICC_TRBE (offsetof(struct acpi_madt_generic_interrupt, \ + trbe_interrupt) + sizeof(u16)) + /* Basic configuration for ACPI */ #ifdef CONFIG_ACPI pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile index 312c164db2ed..05f0ed3ad1d8 100644 --- a/arch/arm64/kernel/Makefile +++ b/arch/arm64/kernel/Makefile @@ -54,6 +54,7 @@ obj-$(CONFIG_PCI) += pci.o obj-$(CONFIG_ARMV8_DEPRECATED) += armv8_deprecated.o obj-$(CONFIG_ACPI) += acpi.o obj-$(CONFIG_ACPI_NUMA) += acpi_numa.o +obj-$(CONFIG_ACPI_TRBE) += acpi_trbe.o obj-$(CONFIG_ARM64_ACPI_PARKING_PROTOCOL) += acpi_parking_protocol.o obj-$(CONFIG_PARAVIRT) += paravirt.o paravirt-spinlocks.o obj-$(CONFIG_PARAVIRT_SPINLOCKS) += paravirt.o paravirt-spinlocks.o diff --git a/arch/arm64/kernel/acpi_trbe.c b/arch/arm64/kernel/acpi_trbe.c new file mode 100644 index 000000000000..5d983ea8c812 --- /dev/null +++ b/arch/arm64/kernel/acpi_trbe.c @@ -0,0 +1,81 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * ACPI probing code for ARM Trace Buffer Extension. + * + * Copyright (C) 2022 ARM Ltd. + */ + +#include <linux/acpi.h> +#include <linux/coresight.h> +#include <linux/platform_device.h> +#include <linux/init.h> + +static struct resource trbe_resources[] = { + { + /* irq */ + .flags = IORESOURCE_IRQ, + } +}; + +static struct platform_device trbe_dev = { + .name = ARMV9_TRBE_PDEV_NAME, + .id = -1, + .resource = trbe_resources, + .num_resources = ARRAY_SIZE(trbe_resources) +}; + +static void arm_trbe_acpi_register_device(void) +{ + int cpu, hetid, irq, ret; + bool first = true; + u16 gsi = 0; + + /* + * Sanity check all the GICC tables for the same interrupt number. + * For now, we only support homogeneous machines. + */ + for_each_possible_cpu(cpu) { + struct acpi_madt_generic_interrupt *gicc; + + gicc = acpi_cpu_get_madt_gicc(cpu); + if (gicc->header.length < ACPI_MADT_GICC_TRBE) + return; + + if (first) { + gsi = gicc->trbe_interrupt; + if (!gsi) + return; + hetid = find_acpi_cpu_topology_hetero_id(cpu); + first = false; + } else if ((gsi != gicc->trbe_interrupt) || + (hetid != find_acpi_cpu_topology_hetero_id(cpu))) { + pr_warn("ACPI: TRBE must be homogeneous\n"); + return; + } + } + + irq = acpi_register_gsi(NULL, gsi, ACPI_LEVEL_SENSITIVE, + ACPI_ACTIVE_HIGH); + if (irq < 0) { + pr_warn("ACPI: TRBE Unable to register interrupt: %d\n", gsi); + return; + } + + trbe_resources[0].start = irq; + ret = platform_device_register(&trbe_dev); + if (ret < 0) { + pr_warn("ACPI: TRBE: Unable to register device\n"); + acpi_unregister_gsi(gsi); + } +} + +static int arm_acpi_trbe_init(void) +{ + if (acpi_disabled) + return 0; + + arm_trbe_acpi_register_device(); + + return 0; +} +device_initcall(arm_acpi_trbe_init) diff --git a/drivers/hwtracing/coresight/Kconfig b/drivers/hwtracing/coresight/Kconfig index 017551d9b89e..efb439fc0a32 100644 --- a/drivers/hwtracing/coresight/Kconfig +++ b/drivers/hwtracing/coresight/Kconfig @@ -200,3 +200,7 @@ config ULTRASOC_SMB called ultrasoc-smb.

endif + +config ACPI_TRBE + depends on ARM64 && ACPI + def_bool y diff --git a/include/linux/coresight.h b/include/linux/coresight.h index 85008a65e21f..fdd2cdaaf898 100644 --- a/include/linux/coresight.h +++ b/include/linux/coresight.h @@ -33,6 +33,8 @@

#define CORESIGHT_UNLOCK 0xc5acce55

+#define ARMV9_TRBE_PDEV_NAME "arm,trbe-v1" + extern struct bus_type coresight_bustype;

enum coresight_dev_type {