| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <3fdc4eb9-a3eb-48c7-be90-6ee305ef4442@vivo.com>
Date: Mon, 28 Jul 2025 20:21:58 +0800
From: Liao Yuanhong <liaoyuanhong@...o.com>
To: Chao Yu <chao@...nel.org>
Cc: linux-f2fs-devel@...ts.sourceforge.net, linux-kernel@...r.kernel.org,
Jaegeuk Kim <jaegeuk@...nel.org>
Subject: Re: [PATCH] f2fs:Add adjustable write endpoints for multiple devices
On 7/28/2025 3:56 PM, Chao Yu wrote:
> On 7/23/25 16:49, Liao Yuanhong wrote:
>> During the development process, we encounter the following two issues:
>>
>> 1.In a multi-device scenario, it's likely that two devices exhibit
>> inconsistent performance, causing fluctuations in performance and making
>> usage and testing inconvenient. Under normal circumstances, we hope to
>> prioritize the use of the device with better performance and consider the
>> device with poorer performance when space is insufficient. Using reuse
>> mode can solve some of the issues, but tests reveal that the fragmentation
>> degree in reuse mode is significantly higher than in default mode, which
>> poses performance risks.
>>
>> 2.We need to examine the differences in data placement for different sizes
>> of storage devices under specific write patterns. Currently, this
>> comparison can only be made by switching storage devices.
>>
>> To address the above issues, I am considering adding a last_secno node. By
>> adjusting this node, we can change the end of the addressing in
>> get_new_segment so that it readdresses from 0 once the set value is
>> reached. The default value of the node is the maximum number of sections
>> for the current storage device, so making no modifications will not affect
>> the current logic. If the space before the set value is already filled with
>> valid data, it will normally write into the free area after the set value.
>>
>> Signed-off-by: Liao Yuanhong <liaoyuanhong@...o.com>
>> ---
>> Documentation/ABI/testing/sysfs-fs-f2fs | 8 ++++++++
>> fs/f2fs/f2fs.h | 1 +
>> fs/f2fs/segment.c | 4 ++--
>> fs/f2fs/super.c | 1 +
>> fs/f2fs/sysfs.c | 9 +++++++++
>> 5 files changed, 21 insertions(+), 2 deletions(-)
>>
>> diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fs
>> index c2a233f2a085..e5ad8dc70cb6 100644
>> --- a/Documentation/ABI/testing/sysfs-fs-f2fs
>> +++ b/Documentation/ABI/testing/sysfs-fs-f2fs
>> @@ -870,3 +870,11 @@ Description: This threshold is used to control triggering garbage collection whi
>> reserved section before preallocating on pinned file.
>> By default, the value is ovp_sections, especially, for zoned ufs, the
>> value is 1.
>> +
>> +What: /sys/fs/f2fs/<disk>/last_secno
>> +Date: July 2025
>> +Contact: "Yuanhong Liao" <liaoyuanhong@...o.com>
>> +Description: This node is used to adjust the addressing end of f2fs when writing.
>> + The default value is the maximum number of sections in the storage. If the
>> + frontend space is full after adjusting the address, it will still address to
>> + the back of the address until reaching the end of the memory.
>> \ No newline at end of file
>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>> index e6dcd7e6f47c..7d93d8671033 100644
>> --- a/fs/f2fs/f2fs.h
>> +++ b/fs/f2fs/f2fs.h
>> @@ -1802,6 +1802,7 @@ 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 last_secno; /* for adjust the end of target device */
>>
>> /* For write statistics */
>> u64 sectors_written_start;
>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
>> index cc82d42ef14c..192619bb2034 100644
>> --- a/fs/f2fs/segment.c
>> +++ b/fs/f2fs/segment.c
>> @@ -2811,7 +2811,7 @@ static int get_new_segment(struct f2fs_sb_info *sbi,
>> secno = find_next_zero_bit(free_i->free_secmap, MAIN_SECS(sbi), hint);
>>
>> #ifdef CONFIG_BLK_DEV_ZONED
>> - if (secno >= MAIN_SECS(sbi) && f2fs_sb_has_blkzoned(sbi)) {
>> + if (secno >= sbi->last_secno && f2fs_sb_has_blkzoned(sbi)) {
>> /* Write only to sequential zones */
>> if (sbi->blkzone_alloc_policy == BLKZONE_ALLOC_ONLY_SEQ) {
>> hint = GET_SEC_FROM_SEG(sbi, sbi->first_seq_zone_segno);
>> @@ -2827,7 +2827,7 @@ static int get_new_segment(struct f2fs_sb_info *sbi,
>> }
>> #endif
>>
>> - if (secno >= MAIN_SECS(sbi)) {
>> + if (secno >= sbi->last_secno) {
>> secno = find_first_zero_bit(free_i->free_secmap,
>> MAIN_SECS(sbi));
>> if (secno >= MAIN_SECS(sbi)) {
>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>> index 30c038413040..5b7de0734da7 100644
>> --- a/fs/f2fs/super.c
>> +++ b/fs/f2fs/super.c
>> @@ -3859,6 +3859,7 @@ static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
>> segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
>> secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
>> total_sections = le32_to_cpu(raw_super->section_count);
>> + sbi->last_secno = total_sections;
>>
>> /* blocks_per_seg should be 512, given the above check */
>> blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
>> diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
>> index bdef926b3377..a4ab5c38e883 100644
>> --- a/fs/f2fs/sysfs.c
>> +++ b/fs/f2fs/sysfs.c
>> @@ -649,6 +649,13 @@ static ssize_t __sbi_store(struct f2fs_attr *a,
>> return count;
>> }
>>
>> + if (!strcmp(a->attr.name, "last_secno")) {
>> + if (t < 0 || t > MAIN_SECS(sbi))
>> + return -EINVAL;
>> + sbi->last_secno = t;
> Will this race w/ f2fs_ioc_resize_fs() which will update MAIN_SECS()?
>
> Thanks,
Yes, I missed the size update that will appear in f2fs_resize_fs. I will
add the modifications in update_fs_metadata() and submit the v2 version
later.
Thanks.
>> + return count;
>> + }
>> +
>> #ifdef CONFIG_F2FS_IOSTAT
>> if (!strcmp(a->attr.name, "iostat_enable")) {
>> sbi->iostat_enable = !!t;
>> @@ -1122,6 +1129,7 @@ F2FS_SBI_GENERAL_RW_ATTR(max_victim_search);
>> F2FS_SBI_GENERAL_RW_ATTR(migration_granularity);
>> F2FS_SBI_GENERAL_RW_ATTR(migration_window_granularity);
>> F2FS_SBI_GENERAL_RW_ATTR(dir_level);
>> +F2FS_SBI_GENERAL_RW_ATTR(last_secno);
>> #ifdef CONFIG_F2FS_IOSTAT
>> F2FS_SBI_GENERAL_RW_ATTR(iostat_enable);
>> F2FS_SBI_GENERAL_RW_ATTR(iostat_period_ms);
>> @@ -1285,6 +1293,7 @@ static struct attribute *f2fs_attrs[] = {
>> ATTR_LIST(discard_idle_interval),
>> ATTR_LIST(gc_idle_interval),
>> ATTR_LIST(umount_discard_timeout),
>> + ATTR_LIST(last_secno),
>> #ifdef CONFIG_F2FS_IOSTAT
>> ATTR_LIST(iostat_enable),
>> ATTR_LIST(iostat_period_ms),
Powered by blists - more mailing lists