All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chao Yu <chao@kernel.org>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Chao Yu <yuchao0@huawei.com>,
	linux-f2fs-devel@lists.sourceforge.net,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/2] f2fs: swap: remove dead codes
Date: Tue, 8 Jun 2021 07:08:50 +0800	[thread overview]
Message-ID: <3272ecd0-f787-7751-3598-19ed8cb1c90f@kernel.org> (raw)
In-Reply-To: <YL5NmfYhLEXOCYiZ@google.com>

On 2021/6/8 0:47, Jaegeuk Kim wrote:
> On 06/07, Chao Yu wrote:
>> On 2021/6/5 7:51, Jaegeuk Kim wrote:
>>> On 06/04, Chao Yu wrote:
>>>> On 2021/5/27 10:12, Chao Yu wrote:
>>>>> On 2021/5/27 9:40, Jaegeuk Kim wrote:
>>>>>> On 05/26, Chao Yu wrote:
>>>>>>> On 2021/5/26 21:30, Jaegeuk Kim wrote:
>>>>>>>> On 05/26, Chao Yu wrote:
>>>>>>>>> After commit af4b6b8edf6a ("f2fs: introduce check_swap_activate_fast()"),
>>>>>>>>> we will never run into original logic of check_swap_activate() before
>>>>>>>>> f2fs supports non 4k-sized page, so let's delete those dead codes.
>>>>>>>>
>>>>>>>> Why not keeping this for large page support in future maybe?
>>>>>>>
>>>>>>> Well, if so, at that time, it would be better to refactor
>>>>>>> check_swap_activate_fast() implementation based on f2fs_map_block() rather
>>>>>>> than refactoring check_swap_activate() implementation based on low efficient
>>>>>>> bmap()?
>>>>>>
>>>>>> Let's first check whether we support large page. Have you quickly tested it?
>>>>>> If we support it now and current flow is just inefficient, I'd say keeping but
>>>>>> refactoring it later.
>>>>>
>>>>> Let me check this, but I guess it missed to handle large page in lots of flows,
>>>>> not sure where we may panic at... :(
>>>>
>>>> I've checked f2fs behavior on 64kb page size in arm64 vm,
>>>> - 'ls -ls' will cause f2fs printing "invalid namelen(0), ino:%u, run fsck
>>>> to fix." each time.
>>>> - after touch a file and umount, fsck reports that image is corrupted.
>>>>
>>>> F2FS uses PAGE_SIZE and F2FS_BLKSIZE directly in a lot of places, I doubt
>>>> most of places may not consider large page size condition.
>>>>
>>>> So, thoughts? I guess we have to admit we do not support 64KB page right
>>>> now... :P
>>>
>>> Yeah, we need to block the large page support unfortunately. :(
>>
>> Could you please queue this series in dev-test?
>>
>> Thanks,
>>
>>>
>>>>
>>>> Thanks,
>>>>
>>>>>
>>>>> Thanks,
>>>>>
>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>>>>>>>>> ---
>>>>>>>>>      fs/f2fs/data.c | 171 +------------------------------------------------
>>>>>>>>>      1 file changed, 3 insertions(+), 168 deletions(-)
>>>>>>>>>
>>>>>>>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>>>>>>>>> index 3058c7e28b11..9c23fde93b76 100644
>>>>>>>>> --- a/fs/f2fs/data.c
>>>>>>>>> +++ b/fs/f2fs/data.c
>>>>>>>>> @@ -3830,67 +3830,7 @@ int f2fs_migrate_page(struct address_space *mapping,
>>>>>>>>>      #endif
>>>>>>>>>      #ifdef CONFIG_SWAP
>>>>>>>>> -static int f2fs_is_file_aligned(struct inode *inode)
>>>>>>>>> -{
>>>>>>>>> -    struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
>>>>>>>>> -    block_t main_blkaddr = SM_I(sbi)->main_blkaddr;
>>>>>>>>> -    block_t cur_lblock;
>>>>>>>>> -    block_t last_lblock;
>>>>>>>>> -    block_t pblock;
>>>>>>>>> -    unsigned long nr_pblocks;
>>>>>>>>> -    unsigned int blocks_per_sec = BLKS_PER_SEC(sbi);
>>>>>>>>> -    unsigned int not_aligned = 0;
>>>>>>>>> -    int ret = 0;
>>>>>>>>> -
>>>>>>>>> -    cur_lblock = 0;
>>>>>>>>> -    last_lblock = bytes_to_blks(inode, i_size_read(inode));
>>>>>>>>> -
>>>>>>>>> -    while (cur_lblock < last_lblock) {
>>>>>>>>> -        struct f2fs_map_blocks map;
>>>>>>>>> -
>>>>>>>>> -        memset(&map, 0, sizeof(map));
>>>>>>>>> -        map.m_lblk = cur_lblock;
>>>>>>>>> -        map.m_len = last_lblock - cur_lblock;
>>>>>>>>> -        map.m_next_pgofs = NULL;
>>>>>>>>> -        map.m_next_extent = NULL;
>>>>>>>>> -        map.m_seg_type = NO_CHECK_TYPE;
>>>>>>>>> -        map.m_may_create = false;
>>>>>>>>> -
>>>>>>>>> -        ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
>>>>>>>>> -        if (ret)
>>>>>>>>> -            goto out;
>>>>>>>>> -
>>>>>>>>> -        /* hole */
>>>>>>>>> -        if (!(map.m_flags & F2FS_MAP_FLAGS)) {
>>>>>>>>> -            f2fs_err(sbi, "Swapfile has holes\n");
>>>>>>>>> -            ret = -ENOENT;
>>>>>>>>> -            goto out;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        pblock = map.m_pblk;
>>>>>>>>> -        nr_pblocks = map.m_len;
>>>>>>>>> -
>>>>>>>>> -        if ((pblock - main_blkaddr) & (blocks_per_sec - 1) ||
>>>>>>>>> -            nr_pblocks & (blocks_per_sec - 1)) {
>>>>>>>>> -            if (f2fs_is_pinned_file(inode)) {
>>>>>>>>> -                f2fs_err(sbi, "Swapfile does not align to section");
>>>>>>>>> -                ret = -EINVAL;
>>>>>>>>> -                goto out;
>>>>>>>>> -            }
>>>>>>>>> -            not_aligned++;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        cur_lblock += nr_pblocks;
>>>>>>>>> -    }
>>>>>>>>> -    if (not_aligned)
>>>>>>>>> -        f2fs_warn(sbi, "Swapfile (%u) is not align to section: \n"
>>>>>>>>> -            "\t1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate()",
>>>>>>>>> -            not_aligned);
>>>>>>>>> -out:
>>>>>>>>> -    return ret;
>>>>>>>>> -}
>>>>>>>>> -
>>>>>>>>> -static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>> +static int check_swap_activate(struct swap_info_struct *sis,
>>>>>>>>>                      struct file *swap_file, sector_t *span)
>>>>>>>>>      {
>>>>>>>>>          struct address_space *mapping = swap_file->f_mapping;
>>>>>>>>> @@ -3907,6 +3847,8 @@ static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>>          unsigned int not_aligned = 0;
>>>>>>>>>          int ret = 0;
>>>>>>>>> +    f2fs_bug_on(sbi, PAGE_SIZE != F2FS_BLKSIZE);
> 
> Let me remove this, since we have this actually.

Looks fine to me.

Thanks,

> 
> in init_f2fs_fs(),
> 
> 	if (PAGE_SIZE != F2FS_BLKSIZE) {
> 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
> 				PAGE_SIZE, F2FS_BLKSIZE);
> 		return -EINVAL;
> 	}
> 
> 
> 
>>>>>>>>> +
>>>>>>>>>          /*
>>>>>>>>>           * Map all the blocks into the extent list.  This code doesn't try
>>>>>>>>>           * to be very smart.
>>>>>>>>> @@ -3986,113 +3928,6 @@ static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>>          return ret;
>>>>>>>>>      }
>>>>>>>>> -/* Copied from generic_swapfile_activate() to check any holes */
>>>>>>>>> -static int check_swap_activate(struct swap_info_struct *sis,
>>>>>>>>> -                struct file *swap_file, sector_t *span)
>>>>>>>>> -{
>>>>>>>>> -    struct address_space *mapping = swap_file->f_mapping;
>>>>>>>>> -    struct inode *inode = mapping->host;
>>>>>>>>> -    struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
>>>>>>>>> -    unsigned blocks_per_page;
>>>>>>>>> -    unsigned long page_no;
>>>>>>>>> -    sector_t probe_block;
>>>>>>>>> -    sector_t last_block;
>>>>>>>>> -    sector_t lowest_block = -1;
>>>>>>>>> -    sector_t highest_block = 0;
>>>>>>>>> -    int nr_extents = 0;
>>>>>>>>> -    int ret = 0;
>>>>>>>>> -
>>>>>>>>> -    if (PAGE_SIZE == F2FS_BLKSIZE)
>>>>>>>>> -        return check_swap_activate_fast(sis, swap_file, span);
>>>>>>>>> -
>>>>>>>>> -    ret = f2fs_is_file_aligned(inode);
>>>>>>>>> -    if (ret)
>>>>>>>>> -        goto out;
>>>>>>>>> -
>>>>>>>>> -    blocks_per_page = bytes_to_blks(inode, PAGE_SIZE);
>>>>>>>>> -
>>>>>>>>> -    /*
>>>>>>>>> -     * Map all the blocks into the extent list.  This code doesn't try
>>>>>>>>> -     * to be very smart.
>>>>>>>>> -     */
>>>>>>>>> -    probe_block = 0;
>>>>>>>>> -    page_no = 0;
>>>>>>>>> -    last_block = bytes_to_blks(inode, i_size_read(inode));
>>>>>>>>> -    while ((probe_block + blocks_per_page) <= last_block &&
>>>>>>>>> -            page_no < sis->max) {
>>>>>>>>> -        unsigned block_in_page;
>>>>>>>>> -        sector_t first_block;
>>>>>>>>> -        sector_t block = 0;
>>>>>>>>> -
>>>>>>>>> -        cond_resched();
>>>>>>>>> -
>>>>>>>>> -        block = probe_block;
>>>>>>>>> -        ret = bmap(inode, &block);
>>>>>>>>> -        if (ret)
>>>>>>>>> -            goto out;
>>>>>>>>> -        if (!block)
>>>>>>>>> -            goto bad_bmap;
>>>>>>>>> -        first_block = block;
>>>>>>>>> -
>>>>>>>>> -        /*
>>>>>>>>> -         * It must be PAGE_SIZE aligned on-disk
>>>>>>>>> -         */
>>>>>>>>> -        if (first_block & (blocks_per_page - 1)) {
>>>>>>>>> -            probe_block++;
>>>>>>>>> -            goto reprobe;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        for (block_in_page = 1; block_in_page < blocks_per_page;
>>>>>>>>> -                    block_in_page++) {
>>>>>>>>> -
>>>>>>>>> -            block = probe_block + block_in_page;
>>>>>>>>> -            ret = bmap(inode, &block);
>>>>>>>>> -            if (ret)
>>>>>>>>> -                goto out;
>>>>>>>>> -            if (!block)
>>>>>>>>> -                goto bad_bmap;
>>>>>>>>> -
>>>>>>>>> -            if (block != first_block + block_in_page) {
>>>>>>>>> -                /* Discontiguity */
>>>>>>>>> -                probe_block++;
>>>>>>>>> -                goto reprobe;
>>>>>>>>> -            }
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        first_block >>= (PAGE_SHIFT - inode->i_blkbits);
>>>>>>>>> -        if (page_no) {    /* exclude the header page */
>>>>>>>>> -            if (first_block < lowest_block)
>>>>>>>>> -                lowest_block = first_block;
>>>>>>>>> -            if (first_block > highest_block)
>>>>>>>>> -                highest_block = first_block;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        /*
>>>>>>>>> -         * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
>>>>>>>>> -         */
>>>>>>>>> -        ret = add_swap_extent(sis, page_no, 1, first_block);
>>>>>>>>> -        if (ret < 0)
>>>>>>>>> -            goto out;
>>>>>>>>> -        nr_extents += ret;
>>>>>>>>> -        page_no++;
>>>>>>>>> -        probe_block += blocks_per_page;
>>>>>>>>> -reprobe:
>>>>>>>>> -        continue;
>>>>>>>>> -    }
>>>>>>>>> -    ret = nr_extents;
>>>>>>>>> -    *span = 1 + highest_block - lowest_block;
>>>>>>>>> -    if (page_no == 0)
>>>>>>>>> -        page_no = 1;    /* force Empty message */
>>>>>>>>> -    sis->max = page_no;
>>>>>>>>> -    sis->pages = page_no - 1;
>>>>>>>>> -    sis->highest_bit = page_no - 1;
>>>>>>>>> -out:
>>>>>>>>> -    return ret;
>>>>>>>>> -bad_bmap:
>>>>>>>>> -    f2fs_err(sbi, "Swapfile has holes\n");
>>>>>>>>> -    return -EINVAL;
>>>>>>>>> -}
>>>>>>>>> -
>>>>>>>>>      static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
>>>>>>>>>                      sector_t *span)
>>>>>>>>>      {
>>>>>>>>> -- 
>>>>>>>>> 2.29.2
>>>>>> .
>>>>>>

WARNING: multiple messages have this Message-ID
From: Chao Yu <chao@kernel.org>
To: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH 1/2] f2fs: swap: remove dead codes
Date: Tue, 8 Jun 2021 07:08:50 +0800	[thread overview]
Message-ID: <3272ecd0-f787-7751-3598-19ed8cb1c90f@kernel.org> (raw)
In-Reply-To: <YL5NmfYhLEXOCYiZ@google.com>

On 2021/6/8 0:47, Jaegeuk Kim wrote:
> On 06/07, Chao Yu wrote:
>> On 2021/6/5 7:51, Jaegeuk Kim wrote:
>>> On 06/04, Chao Yu wrote:
>>>> On 2021/5/27 10:12, Chao Yu wrote:
>>>>> On 2021/5/27 9:40, Jaegeuk Kim wrote:
>>>>>> On 05/26, Chao Yu wrote:
>>>>>>> On 2021/5/26 21:30, Jaegeuk Kim wrote:
>>>>>>>> On 05/26, Chao Yu wrote:
>>>>>>>>> After commit af4b6b8edf6a ("f2fs: introduce check_swap_activate_fast()"),
>>>>>>>>> we will never run into original logic of check_swap_activate() before
>>>>>>>>> f2fs supports non 4k-sized page, so let's delete those dead codes.
>>>>>>>>
>>>>>>>> Why not keeping this for large page support in future maybe?
>>>>>>>
>>>>>>> Well, if so, at that time, it would be better to refactor
>>>>>>> check_swap_activate_fast() implementation based on f2fs_map_block() rather
>>>>>>> than refactoring check_swap_activate() implementation based on low efficient
>>>>>>> bmap()?
>>>>>>
>>>>>> Let's first check whether we support large page. Have you quickly tested it?
>>>>>> If we support it now and current flow is just inefficient, I'd say keeping but
>>>>>> refactoring it later.
>>>>>
>>>>> Let me check this, but I guess it missed to handle large page in lots of flows,
>>>>> not sure where we may panic at... :(
>>>>
>>>> I've checked f2fs behavior on 64kb page size in arm64 vm,
>>>> - 'ls -ls' will cause f2fs printing "invalid namelen(0), ino:%u, run fsck
>>>> to fix." each time.
>>>> - after touch a file and umount, fsck reports that image is corrupted.
>>>>
>>>> F2FS uses PAGE_SIZE and F2FS_BLKSIZE directly in a lot of places, I doubt
>>>> most of places may not consider large page size condition.
>>>>
>>>> So, thoughts? I guess we have to admit we do not support 64KB page right
>>>> now... :P
>>>
>>> Yeah, we need to block the large page support unfortunately. :(
>>
>> Could you please queue this series in dev-test?
>>
>> Thanks,
>>
>>>
>>>>
>>>> Thanks,
>>>>
>>>>>
>>>>> Thanks,
>>>>>
>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>>>>>>>>> ---
>>>>>>>>>      fs/f2fs/data.c | 171 +------------------------------------------------
>>>>>>>>>      1 file changed, 3 insertions(+), 168 deletions(-)
>>>>>>>>>
>>>>>>>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
>>>>>>>>> index 3058c7e28b11..9c23fde93b76 100644
>>>>>>>>> --- a/fs/f2fs/data.c
>>>>>>>>> +++ b/fs/f2fs/data.c
>>>>>>>>> @@ -3830,67 +3830,7 @@ int f2fs_migrate_page(struct address_space *mapping,
>>>>>>>>>      #endif
>>>>>>>>>      #ifdef CONFIG_SWAP
>>>>>>>>> -static int f2fs_is_file_aligned(struct inode *inode)
>>>>>>>>> -{
>>>>>>>>> -    struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
>>>>>>>>> -    block_t main_blkaddr = SM_I(sbi)->main_blkaddr;
>>>>>>>>> -    block_t cur_lblock;
>>>>>>>>> -    block_t last_lblock;
>>>>>>>>> -    block_t pblock;
>>>>>>>>> -    unsigned long nr_pblocks;
>>>>>>>>> -    unsigned int blocks_per_sec = BLKS_PER_SEC(sbi);
>>>>>>>>> -    unsigned int not_aligned = 0;
>>>>>>>>> -    int ret = 0;
>>>>>>>>> -
>>>>>>>>> -    cur_lblock = 0;
>>>>>>>>> -    last_lblock = bytes_to_blks(inode, i_size_read(inode));
>>>>>>>>> -
>>>>>>>>> -    while (cur_lblock < last_lblock) {
>>>>>>>>> -        struct f2fs_map_blocks map;
>>>>>>>>> -
>>>>>>>>> -        memset(&map, 0, sizeof(map));
>>>>>>>>> -        map.m_lblk = cur_lblock;
>>>>>>>>> -        map.m_len = last_lblock - cur_lblock;
>>>>>>>>> -        map.m_next_pgofs = NULL;
>>>>>>>>> -        map.m_next_extent = NULL;
>>>>>>>>> -        map.m_seg_type = NO_CHECK_TYPE;
>>>>>>>>> -        map.m_may_create = false;
>>>>>>>>> -
>>>>>>>>> -        ret = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_FIEMAP);
>>>>>>>>> -        if (ret)
>>>>>>>>> -            goto out;
>>>>>>>>> -
>>>>>>>>> -        /* hole */
>>>>>>>>> -        if (!(map.m_flags & F2FS_MAP_FLAGS)) {
>>>>>>>>> -            f2fs_err(sbi, "Swapfile has holes\n");
>>>>>>>>> -            ret = -ENOENT;
>>>>>>>>> -            goto out;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        pblock = map.m_pblk;
>>>>>>>>> -        nr_pblocks = map.m_len;
>>>>>>>>> -
>>>>>>>>> -        if ((pblock - main_blkaddr) & (blocks_per_sec - 1) ||
>>>>>>>>> -            nr_pblocks & (blocks_per_sec - 1)) {
>>>>>>>>> -            if (f2fs_is_pinned_file(inode)) {
>>>>>>>>> -                f2fs_err(sbi, "Swapfile does not align to section");
>>>>>>>>> -                ret = -EINVAL;
>>>>>>>>> -                goto out;
>>>>>>>>> -            }
>>>>>>>>> -            not_aligned++;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        cur_lblock += nr_pblocks;
>>>>>>>>> -    }
>>>>>>>>> -    if (not_aligned)
>>>>>>>>> -        f2fs_warn(sbi, "Swapfile (%u) is not align to section: \n"
>>>>>>>>> -            "\t1) creat(), 2) ioctl(F2FS_IOC_SET_PIN_FILE), 3) fallocate()",
>>>>>>>>> -            not_aligned);
>>>>>>>>> -out:
>>>>>>>>> -    return ret;
>>>>>>>>> -}
>>>>>>>>> -
>>>>>>>>> -static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>> +static int check_swap_activate(struct swap_info_struct *sis,
>>>>>>>>>                      struct file *swap_file, sector_t *span)
>>>>>>>>>      {
>>>>>>>>>          struct address_space *mapping = swap_file->f_mapping;
>>>>>>>>> @@ -3907,6 +3847,8 @@ static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>>          unsigned int not_aligned = 0;
>>>>>>>>>          int ret = 0;
>>>>>>>>> +    f2fs_bug_on(sbi, PAGE_SIZE != F2FS_BLKSIZE);
> 
> Let me remove this, since we have this actually.

Looks fine to me.

Thanks,

> 
> in init_f2fs_fs(),
> 
> 	if (PAGE_SIZE != F2FS_BLKSIZE) {
> 		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
> 				PAGE_SIZE, F2FS_BLKSIZE);
> 		return -EINVAL;
> 	}
> 
> 
> 
>>>>>>>>> +
>>>>>>>>>          /*
>>>>>>>>>           * Map all the blocks into the extent list.  This code doesn't try
>>>>>>>>>           * to be very smart.
>>>>>>>>> @@ -3986,113 +3928,6 @@ static int check_swap_activate_fast(struct swap_info_struct *sis,
>>>>>>>>>          return ret;
>>>>>>>>>      }
>>>>>>>>> -/* Copied from generic_swapfile_activate() to check any holes */
>>>>>>>>> -static int check_swap_activate(struct swap_info_struct *sis,
>>>>>>>>> -                struct file *swap_file, sector_t *span)
>>>>>>>>> -{
>>>>>>>>> -    struct address_space *mapping = swap_file->f_mapping;
>>>>>>>>> -    struct inode *inode = mapping->host;
>>>>>>>>> -    struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
>>>>>>>>> -    unsigned blocks_per_page;
>>>>>>>>> -    unsigned long page_no;
>>>>>>>>> -    sector_t probe_block;
>>>>>>>>> -    sector_t last_block;
>>>>>>>>> -    sector_t lowest_block = -1;
>>>>>>>>> -    sector_t highest_block = 0;
>>>>>>>>> -    int nr_extents = 0;
>>>>>>>>> -    int ret = 0;
>>>>>>>>> -
>>>>>>>>> -    if (PAGE_SIZE == F2FS_BLKSIZE)
>>>>>>>>> -        return check_swap_activate_fast(sis, swap_file, span);
>>>>>>>>> -
>>>>>>>>> -    ret = f2fs_is_file_aligned(inode);
>>>>>>>>> -    if (ret)
>>>>>>>>> -        goto out;
>>>>>>>>> -
>>>>>>>>> -    blocks_per_page = bytes_to_blks(inode, PAGE_SIZE);
>>>>>>>>> -
>>>>>>>>> -    /*
>>>>>>>>> -     * Map all the blocks into the extent list.  This code doesn't try
>>>>>>>>> -     * to be very smart.
>>>>>>>>> -     */
>>>>>>>>> -    probe_block = 0;
>>>>>>>>> -    page_no = 0;
>>>>>>>>> -    last_block = bytes_to_blks(inode, i_size_read(inode));
>>>>>>>>> -    while ((probe_block + blocks_per_page) <= last_block &&
>>>>>>>>> -            page_no < sis->max) {
>>>>>>>>> -        unsigned block_in_page;
>>>>>>>>> -        sector_t first_block;
>>>>>>>>> -        sector_t block = 0;
>>>>>>>>> -
>>>>>>>>> -        cond_resched();
>>>>>>>>> -
>>>>>>>>> -        block = probe_block;
>>>>>>>>> -        ret = bmap(inode, &block);
>>>>>>>>> -        if (ret)
>>>>>>>>> -            goto out;
>>>>>>>>> -        if (!block)
>>>>>>>>> -            goto bad_bmap;
>>>>>>>>> -        first_block = block;
>>>>>>>>> -
>>>>>>>>> -        /*
>>>>>>>>> -         * It must be PAGE_SIZE aligned on-disk
>>>>>>>>> -         */
>>>>>>>>> -        if (first_block & (blocks_per_page - 1)) {
>>>>>>>>> -            probe_block++;
>>>>>>>>> -            goto reprobe;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        for (block_in_page = 1; block_in_page < blocks_per_page;
>>>>>>>>> -                    block_in_page++) {
>>>>>>>>> -
>>>>>>>>> -            block = probe_block + block_in_page;
>>>>>>>>> -            ret = bmap(inode, &block);
>>>>>>>>> -            if (ret)
>>>>>>>>> -                goto out;
>>>>>>>>> -            if (!block)
>>>>>>>>> -                goto bad_bmap;
>>>>>>>>> -
>>>>>>>>> -            if (block != first_block + block_in_page) {
>>>>>>>>> -                /* Discontiguity */
>>>>>>>>> -                probe_block++;
>>>>>>>>> -                goto reprobe;
>>>>>>>>> -            }
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        first_block >>= (PAGE_SHIFT - inode->i_blkbits);
>>>>>>>>> -        if (page_no) {    /* exclude the header page */
>>>>>>>>> -            if (first_block < lowest_block)
>>>>>>>>> -                lowest_block = first_block;
>>>>>>>>> -            if (first_block > highest_block)
>>>>>>>>> -                highest_block = first_block;
>>>>>>>>> -        }
>>>>>>>>> -
>>>>>>>>> -        /*
>>>>>>>>> -         * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks
>>>>>>>>> -         */
>>>>>>>>> -        ret = add_swap_extent(sis, page_no, 1, first_block);
>>>>>>>>> -        if (ret < 0)
>>>>>>>>> -            goto out;
>>>>>>>>> -        nr_extents += ret;
>>>>>>>>> -        page_no++;
>>>>>>>>> -        probe_block += blocks_per_page;
>>>>>>>>> -reprobe:
>>>>>>>>> -        continue;
>>>>>>>>> -    }
>>>>>>>>> -    ret = nr_extents;
>>>>>>>>> -    *span = 1 + highest_block - lowest_block;
>>>>>>>>> -    if (page_no == 0)
>>>>>>>>> -        page_no = 1;    /* force Empty message */
>>>>>>>>> -    sis->max = page_no;
>>>>>>>>> -    sis->pages = page_no - 1;
>>>>>>>>> -    sis->highest_bit = page_no - 1;
>>>>>>>>> -out:
>>>>>>>>> -    return ret;
>>>>>>>>> -bad_bmap:
>>>>>>>>> -    f2fs_err(sbi, "Swapfile has holes\n");
>>>>>>>>> -    return -EINVAL;
>>>>>>>>> -}
>>>>>>>>> -
>>>>>>>>>      static int f2fs_swap_activate(struct swap_info_struct *sis, struct file *file,
>>>>>>>>>                      sector_t *span)
>>>>>>>>>      {
>>>>>>>>> -- 
>>>>>>>>> 2.29.2
>>>>>> .
>>>>>>


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

  reply	other threads:[~2021-06-07 23:08 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-26  6:29 Chao Yu
2021-05-26  6:29 ` [f2fs-dev] " Chao Yu
2021-05-26  6:29 ` [PATCH 2/2] f2fs: swap: support migrating swapfile in aligned write mode Chao Yu
2021-05-26  6:29   ` [f2fs-dev] " Chao Yu
2021-05-26 13:30 ` [PATCH 1/2] f2fs: swap: remove dead codes Jaegeuk Kim
2021-05-26 13:30   ` [f2fs-dev] " Jaegeuk Kim
2021-05-26 15:02   ` Chao Yu
2021-05-26 15:02     ` [f2fs-dev] " Chao Yu
2021-05-27  1:40     ` Jaegeuk Kim
2021-05-27  1:40       ` [f2fs-dev] " Jaegeuk Kim
2021-05-27  2:12       ` Chao Yu
2021-05-27  2:12         ` [f2fs-dev] " Chao Yu
2021-06-04 11:12         ` Chao Yu
2021-06-04 11:12           ` [f2fs-dev] " Chao Yu
2021-06-04 23:51           ` Jaegeuk Kim
2021-06-04 23:51             ` [f2fs-dev] " Jaegeuk Kim
2021-06-07 15:14             ` Chao Yu
2021-06-07 15:14               ` [f2fs-dev] " Chao Yu
2021-06-07 16:47               ` Jaegeuk Kim
2021-06-07 16:47                 ` [f2fs-dev] " Jaegeuk Kim
2021-06-07 23:08                 ` Chao Yu [this message]
2021-06-07 23:08                   ` Chao Yu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=3272ecd0-f787-7751-3598-19ed8cb1c90f@kernel.org \
    --to=chao@kernel.org \
    --cc=jaegeuk@kernel.org \
    --cc=linux-f2fs-devel@lists.sourceforge.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=yuchao0@huawei.com \
    --subject='Re: [PATCH 1/2] f2fs: swap: remove dead codes' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.