From: Bixuan Cui cuibixuan@huawei.com

ascend inclusion category: feature bugzilla: 48163 CVE: NA

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

After the MAP_LOCKED flag is enabled, resource contention occurs when multiple tasks map/munmap memory at the same time, causing performance loss.

Add sysctl_share_pool_map_lock_enable to control whether the map is locked in sp_mmap().

Signed-off-by: Bixuan Cui cuibixuan@huawei.com Reviewed-by: Ding Tianhong dingtianhong@huawei.com Reviewed-by: Hanjun Guo guohanjun@huawei.com Signed-off-by: Yang Yingliang yangyingliang@huawei.com --- include/linux/share_pool.h | 2 ++ kernel/sysctl.c | 10 ++++++++++ mm/share_pool.c | 12 ++++++++++-- 3 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/include/linux/share_pool.h b/include/linux/share_pool.h index fb7237351a995..c3120b7b24948 100644 --- a/include/linux/share_pool.h +++ b/include/linux/share_pool.h @@ -38,6 +38,8 @@ extern int sysctl_sp_debug_mode;

extern int enable_ascend_share_pool;

+extern int sysctl_share_pool_map_lock_enable; + #ifdef CONFIG_HAVE_ARCH_HUGE_VMALLOC extern bool vmap_allow_huge; #endif diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 12eb6c769afcc..4d8d7c15a6ebc 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -1265,6 +1265,16 @@ static struct ctl_table kern_table[] = { .extra1 = &zero, .extra2 = &one, }, + { + /* 0: map_unlock, 1: map_lock */ + .procname = "share_pool_map_lock_enable", + .data = &sysctl_share_pool_map_lock_enable, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec_minmax, + .extra1 = &zero, + .extra2 = &one, + }, #endif { } }; diff --git a/mm/share_pool.c b/mm/share_pool.c index e6feaad8bdd13..4fa693bd7f973 100644 --- a/mm/share_pool.c +++ b/mm/share_pool.c @@ -69,6 +69,8 @@ int sysctl_ac_mode = AC_NONE; /* debug mode */ int sysctl_sp_debug_mode;

+int sysctl_share_pool_map_lock_enable; + /* idr of all sp_groups */ static DEFINE_IDR(sp_group_idr);

@@ -1227,11 +1229,17 @@ static unsigned long sp_mmap(struct mm_struct *mm, struct file *file, unsigned long addr = spa->va_start; unsigned long size = spa_size(spa); unsigned long prot = PROT_READ | PROT_WRITE; - unsigned long flags = MAP_FIXED | MAP_SHARED | MAP_LOCKED | - MAP_POPULATE | MAP_SHARE_POOL; + unsigned long flags = MAP_FIXED | MAP_SHARED | MAP_POPULATE | + MAP_SHARE_POOL; unsigned long vm_flags = VM_NORESERVE | VM_SHARE_POOL | VM_DONTCOPY; unsigned long pgoff = (addr - MMAP_SHARE_POOL_START) >> PAGE_SHIFT;

+ /* Mark the mapped region to be locked. After the MAP_LOCKED is enable, + * multiple tasks will preempt resources, causing performance loss. + */ + if (sysctl_share_pool_map_lock_enable) + flags |= MAP_LOCKED; + atomic_inc(&spa->use_count); addr = __do_mmap(mm, file, addr, size, prot, flags, vm_flags, pgoff, populate, NULL);

From: "Martin K. Petersen" martin.petersen@oracle.com

hulk inclusion category: bugfix bugzilla: 46833 CVE: NA

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

Fix: https://gitee.com/src-openeuler/util-linux/issues/I28N07

Origin patch: https://patchwork.kernel.org/project/linux-scsi/patch/20190227041941. 1568-1-martin.petersen@oracle.com/

If partition table changed online, will record partition read-only flag.

Some devices come online in write protected state and switch to read-write once they are ready to process I/O requests. These devices broke with commit 20bd1d026aac ("scsi: sd: Keep disk read-only when re-reading partition") because we had no way to distinguish between a user decision to set a block_device read-only and the actual hardware device being write-protected.

Because partitions are dropped and recreated on revalidate we are unable to persist any user-provided policy in hd_struct. Introduce a bitmap in struct gendisk to track the user configuration. This bitmap is updated when BLKROSET is called on a given disk or partition.

A helper function, get_user_ro(), is provided to determine whether the ioctl has forced read-only state for a given block device. This helper is used by set_disk_ro() and add_partition() to ensure that both existing and newly created partitions will get the correct state.

- If BLKROSET sets a whole disk device read-only, all partitions will now end up in a read-only state.

- If BLKROSET sets a given partition read-only, that partition will remain read-only post revalidate.

- Otherwise both the whole disk device and any partitions will reflect the write protect state of the underlying device.

Since nobody knows what "policy" means, rename the field to "read_only" for clarity.

Cc: Jeremy Cline jeremy@jcline.org Cc: Oleksii Kurochko olkuroch@cisco.com Cc: stable@vger.kernel.org # v4.16+ Reported-by: Oleksii Kurochko olkuroch@cisco.com Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=201221 Fixes: 20bd1d026aac ("scsi: sd: Keep disk read-only when re-reading partition") Signed-off-by: Martin K. Petersen martin.petersen@oracle.com Signed-off-by: Ye Bin yebin10@huawei.com Reviewed-by: Jason Yan yanaijie@huawei.com Signed-off-by: Yang Yingliang yangyingliang@huawei.com --- block/blk-core.c | 2 +- block/genhd.c | 34 ++++++++++++++++++++++++---------- block/ioctl.c | 4 ++++ block/partition-generic.c | 7 +++++-- include/linux/genhd.h | 11 +++++++---- 5 files changed, 41 insertions(+), 17 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c index 528c53592aa2e..93970be37c17c 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -2171,7 +2171,7 @@ static inline bool bio_check_ro(struct bio *bio, struct hd_struct *part) { const int op = bio_op(bio);

- if (part->policy && op_is_write(op)) { + if (part->read_only && op_is_write(op)) { char b[BDEVNAME_SIZE];

if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) diff --git a/block/genhd.c b/block/genhd.c index d8a9c901eaefd..d9ec6bb4f880a 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -1543,26 +1543,40 @@ static void set_disk_ro_uevent(struct gendisk *gd, int ro) kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); }

-void set_device_ro(struct block_device *bdev, int flag) +void set_device_ro(struct block_device *bdev, bool state) { - bdev->bd_part->policy = flag; + bdev->bd_part->read_only = state; }

EXPORT_SYMBOL(set_device_ro);

-void set_disk_ro(struct gendisk *disk, int flag) +bool get_user_ro(struct gendisk *disk, unsigned int partno) +{ + /* Is the user read-only bit set for the whole disk device? */ + if (test_bit(0, disk->user_ro_bitmap)) + return true; + + /* Is the user read-only bit set for this particular partition? */ + if (test_bit(partno, disk->user_ro_bitmap)) + return true; + + return false; +} + +void set_disk_ro(struct gendisk *disk, bool state) { struct disk_part_iter piter; struct hd_struct *part;

- if (disk->part0.policy != flag) { - set_disk_ro_uevent(disk, flag); - disk->part0.policy = flag; - } + if (disk->part0.read_only != state) + set_disk_ro_uevent(disk, state);

- disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); + disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY_PART0); while ((part = disk_part_iter_next(&piter))) - part->policy = flag; + if (get_user_ro(disk, part->partno)) + part->read_only = true; + else + part->read_only = state; disk_part_iter_exit(&piter); }

@@ -1572,7 +1586,7 @@ int bdev_read_only(struct block_device *bdev) { if (!bdev) return 0; - return bdev->bd_part->policy; + return bdev->bd_part->read_only; }

EXPORT_SYMBOL(bdev_read_only); diff --git a/block/ioctl.c b/block/ioctl.c index 5a6157b3735a0..899ffd50a7c6b 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -455,6 +455,10 @@ static int blkdev_roset(struct block_device *bdev, fmode_t mode, return ret; if (get_user(n, (int __user *)arg)) return -EFAULT; + if (n) + set_bit(bdev->bd_partno, bdev->bd_disk->user_ro_bitmap); + else + clear_bit(bdev->bd_partno, bdev->bd_disk->user_ro_bitmap); set_device_ro(bdev, n); return 0; } diff --git a/block/partition-generic.c b/block/partition-generic.c index d86d794d28e1a..63b82df5bbb40 100644 --- a/block/partition-generic.c +++ b/block/partition-generic.c @@ -98,7 +98,7 @@ static ssize_t part_ro_show(struct device *dev, struct device_attribute *attr, char *buf) { struct hd_struct *p = dev_to_part(dev); - return sprintf(buf, "%d\n", p->policy ? 1 : 0); + return sprintf(buf, "%u\n", p->read_only ? 1 : 0); }

static ssize_t part_alignment_offset_show(struct device *dev, @@ -352,7 +352,10 @@ struct hd_struct *add_partition(struct gendisk *disk, int partno, queue_limit_discard_alignment(&disk->queue->limits, start); p->nr_sects = len; p->partno = partno; - p->policy = get_disk_ro(disk); + if (get_user_ro(disk, partno)) + p->read_only = true; + else + p->read_only = get_disk_ro(disk); p->disk = disk;

if (info) { diff --git a/include/linux/genhd.h b/include/linux/genhd.h index 1a4cccbbec00c..f65b200e28846 100644 --- a/include/linux/genhd.h +++ b/include/linux/genhd.h @@ -116,7 +116,8 @@ struct hd_struct { unsigned int discard_alignment; struct device __dev; struct kobject *holder_dir; - int policy, partno; + bool read_only; + int partno; struct partition_meta_info *info; #ifdef CONFIG_FAIL_MAKE_REQUEST int make_it_fail; @@ -202,6 +203,7 @@ struct gendisk { */ struct disk_part_tbl __rcu *part_tbl; struct hd_struct part0; + DECLARE_BITMAP(user_ro_bitmap, DISK_MAX_PARTS);

const struct block_device_operations *fops; struct request_queue *queue; @@ -440,12 +442,13 @@ extern void del_gendisk(struct gendisk *gp); extern struct gendisk *get_gendisk(dev_t dev, int *partno); extern struct block_device *bdget_disk(struct gendisk *disk, int partno);

-extern void set_device_ro(struct block_device *bdev, int flag); -extern void set_disk_ro(struct gendisk *disk, int flag); +extern void set_device_ro(struct block_device *bdev, bool state); +extern void set_disk_ro(struct gendisk *disk, bool state); +extern bool get_user_ro(struct gendisk *disk, unsigned int partno);

static inline int get_disk_ro(struct gendisk *disk) { - return disk->part0.policy; + return disk->part0.read_only; }

extern void disk_block_events(struct gendisk *disk);