On 8/8/2025 4:51 PM, Chao Yu wrote:
On 8/8/2025 3:29 PM, Liao Yuanhong wrote:Currently, we have encountered some issues while testing ZUFS. In situations near the storage limit (e.g., 50GB remaining), and aftersimulating fragmentation by repeatedly writing and deleting data, we found that application installation and startup tests conducted after idling for a few minutes take significantly longer several times that of traditional UFS. Tracing the operations revealed that the majority of I/Os were issuedby background GC, which blocks normal I/O operations.Under normal circumstances, ZUFS indeed requires more background GC and employs a more aggressive GC strategy. However, I aim to find a way to minimize the impact on regular I/O operations under these near-limit conditions. To address this, I have introduced a bggc_block_io feature,which controls the prioritization of background GC in the presence of I/Os. This switch can be adjusted at the framework level to implement differentstrategies. If set to ALL_IO_PRIOR, all background GC operations will beskipped during active I/O issuance. The default option remains consistentwith the current strategy, ensuring no change in behavior. Signed-off-by: Liao Yuanhong <liaoyuanh...@vivo.com> --- Changes in v4:Adjust the default policy ALL_IO_PRIOR to 0 and modify the description tomatch this change Changes in v3: Modified the issue where it does not work after closing CONFIG_BLK_DEV_ZONED. Changes in v2: Non ZUFS can also be adjusted through this option. --- Documentation/ABI/testing/sysfs-fs-f2fs | 13 +++++++++++++ fs/f2fs/f2fs.h | 18 +++++++++++------- fs/f2fs/super.c | 2 ++ fs/f2fs/sysfs.c | 9 +++++++++ 4 files changed, 35 insertions(+), 7 deletions(-)diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fsindex bc0e7fefc39d..21e6951919de 100644 --- a/Documentation/ABI/testing/sysfs-fs-f2fs +++ b/Documentation/ABI/testing/sysfs-fs-f2fs @@ -883,3 +883,16 @@ Date: June 2025 Contact: "Daeho Jeong" <daehoje...@google.com>Description: Control GC algorithm for boost GC. 0: cost benefit, 1: greedyDefault: 1 + +What: /sys/fs/f2fs/<disk>/bggc_block_io +Date: August 2025 +Contact: "Liao Yuanhong" <liaoyuanh...@vivo.com>+Description: Used to adjust the BG_GC priority when issuing IO, with a default value+ of 0. Specifically, for ZUFS, the default value is 1. ++ ================== =============================================+ value description + bggc_block_io = 0 Stop background GC when issuing I/O+ bggc_block_io = 1 Stop background GC only when issuing read I/O+ bggc_block_io = 2 Prioritize background GC+ ================== =============================================diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 46be7560548c..440542ea0646 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -155,6 +155,12 @@ enum blkzone_allocation_policy {BLKZONE_ALLOC_PRIOR_CONV, /* Prioritize writing to conventional zones */}; +enum bggc_block_io_policy { + ALL_IO_PRIOR, + READ_IO_PRIOR, + BGGC_PRIOR,Hi Yuanhong, Just nitpick, :) A little bit confuse for BGGC_PRIOR naming, since it won't submit bggc IO in prior to other IO, just not be aware other IOs. So, what do you think of renaming as below? Keeping align w/ naming of discard_policy.io_aware. /sys/fs/f2fs/<dev>/bggc_io_aware enum bggc_io_aware_policy {AWARE_ALL_IO, /* skip background GC if there is any kind of pending IO */ AWARE_READ_IO, /* skip background GC if there is pending read IO */AWARE_NONE, /* don't aware IO for background GC */ }; Thanks,
Thank you for your suggestion. I will submit a revised new version. Thanks, Liao
+}; + /** An implementation of an rwsem that is explicitly unfair to readers. This * prevents priority inversion when a low-priority reader acquires the read lock@@ -1804,6 +1810,7 @@ struct f2fs_sb_info { spinlock_t dev_lock; /* protect dirty_device */bool aligned_blksize; /* all devices has the same logical blksize */ unsigned int first_seq_zone_segno; /* first segno in sequential zone */ + unsigned int bggc_block_io; /* For adjust the BG_GC priority when issuing IO *//* For write statistics */ u64 sectors_written_start;@@ -2998,13 +3005,10 @@ static inline bool is_idle(struct f2fs_sb_info *sbi, int type)if (sbi->gc_mode == GC_URGENT_HIGH) return true; - if (zoned_gc) { - if (is_inflight_read_io(sbi)) - return false; - } else { - if (is_inflight_io(sbi, type)) - return false; - }+ if (sbi->bggc_block_io == READ_IO_PRIOR && is_inflight_read_io(sbi))+ return false;+ if (sbi->bggc_block_io == ALL_IO_PRIOR && is_inflight_io(sbi, type))+ return false; if (sbi->gc_mode == GC_URGENT_MID) return true; diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index e16c4e2830c2..a21cecc5424e 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c@@ -4629,9 +4629,11 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); sbi->aligned_blksize = true; + sbi->bggc_block_io = ALL_IO_PRIOR; #ifdef CONFIG_BLK_DEV_ZONED sbi->max_open_zones = UINT_MAX; sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ; + sbi->bggc_block_io = READ_IO_PRIOR; #endif for (i = 0; i < max_devices; i++) { diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c index f736052dea50..1b587908f049 100644 --- a/fs/f2fs/sysfs.c +++ b/fs/f2fs/sysfs.c @@ -866,6 +866,13 @@ static ssize_t __sbi_store(struct f2fs_attr *a, return count; } + if (!strcmp(a->attr.name, "bggc_block_io")) { + if (t < ALL_IO_PRIOR || t > BGGC_PRIOR) + return -EINVAL; + sbi->bggc_block_io = t; + return count; + } + *ui = (unsigned int)t; return count; @@ -1175,6 +1182,7 @@ F2FS_SBI_GENERAL_RW_ATTR(blkzone_alloc_policy); #endif F2FS_SBI_GENERAL_RW_ATTR(carve_out); F2FS_SBI_GENERAL_RW_ATTR(reserved_pin_section); +F2FS_SBI_GENERAL_RW_ATTR(bggc_block_io); /* STAT_INFO ATTR */ #ifdef CONFIG_F2FS_STAT_FS @@ -1303,6 +1311,7 @@ static struct attribute *f2fs_attrs[] = { ATTR_LIST(discard_idle_interval), ATTR_LIST(gc_idle_interval), ATTR_LIST(umount_discard_timeout), + ATTR_LIST(bggc_block_io), #ifdef CONFIG_F2FS_IOSTAT ATTR_LIST(iostat_enable), ATTR_LIST(iostat_period_ms),
_______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
