Re: [PATCH v2] fat: Batched discard support for fat
From: Lukas Czerner
Date: Thu Feb 24 2011 - 06:04:04 EST
On Thu, 24 Feb 2011, Kyungmin Park wrote:
> On Thu, Feb 24, 2011 at 5:53 PM, Lukas Czerner <lczerner@xxxxxxxxxx> wrote:
> > On Thu, 24 Feb 2011, Kyungmin Park wrote:
> >
> >> From: Kyungmin Park <kyungmin.park@xxxxxxxxxxx>
> >>
> >> FAT supports batched discard as ext4.
> >>
> >> Cited from Lukas words.
> >> "The current solution is not ideal because of its bad performance impact.
> >> So basic idea to improve things is to avoid discarding every time some
> >> blocks are freed. and instead batching is together into bigger trims,
> >> which tends to be more effective."
> >>
> >> You can find an information in detail at following URLs.
> >> http://lwn.net/Articles/397538/
> >> http://lwn.net/Articles/383933/
> >>
> >> Signed-off-by: Kyungmin Park <kyungmin.park@xxxxxxxxxxx>
> >
> > Hi Kyungmin,
> > thanks to second version. How did you test it ? You can try xtestest
> > 251 for simple verification.
> > I can not really comment on FAT specific code, however I have couple of
> > comments bellow.
> >
> > Thanks!
> > -Lukas
> >
> >> ---
> >> Changelog V2:
> >> Use the given start and len as Lukas comments
> >> Check the queue supports discard feature
> >> ---
> >> diff --git a/fs/fat/fat.h b/fs/fat/fat.h
> >> index f504089..08b53e1 100644
> >> --- a/fs/fat/fat.h
> >> +++ b/fs/fat/fat.h
> >> @@ -299,6 +299,7 @@ extern int fat_alloc_clusters(struct inode *inode, int *cluster,
> >> int nr_cluster);
> >> extern int fat_free_clusters(struct inode *inode, int cluster);
> >> extern int fat_count_free_clusters(struct super_block *sb);
> >> +extern int fat_trim_fs(struct super_block *sb, struct fstrim_range *range);
> >>
> >> /* fat/file.c */
> >> extern long fat_generic_ioctl(struct file *filp, unsigned int cmd,
> >> diff --git a/fs/fat/fatent.c b/fs/fat/fatent.c
> >> index b47d2c9..ea31ee4 100644
> >> --- a/fs/fat/fatent.c
> >> +++ b/fs/fat/fatent.c
> >> @@ -1,6 +1,8 @@
> >> /*
> >> * Copyright (C) 2004, OGAWA Hirofumi
> >> * Released under GPL v2.
> >> + *
> >> + * Batched discard support by Kyungmin Park <kyungmin.park@xxxxxxxxxxx>
> >> */
> >>
> >> #include <linux/module.h>
> >> @@ -541,6 +543,16 @@ out:
> >> return err;
> >> }
> >>
> >> +static int fat_issue_discard(struct super_block *sb, int cluster, int nr_clus)
> >> +{
> >> + struct msdos_sb_info *sbi = MSDOS_SB(sb);
> >> + sector_t block, nr_blocks;
> >> +
> >> + block = fat_clus_to_blknr(sbi, cluster);
> >> + nr_blocks = nr_clus * sbi->sec_per_clus;
> >> + return sb_issue_discard(sb, block, nr_blocks, GFP_NOFS, 0);
> > ^^^^^^^^
> > This patch does not pass checkpatch.pl script. Use tabs for indention.
> right, I will fix it. now passed the checkpatch.pl.
>
> >
> >> +}
> >> +
> >> int fat_free_clusters(struct inode *inode, int cluster)
> >> {
> >> struct super_block *sb = inode->i_sb;
> >> @@ -575,11 +587,7 @@ int fat_free_clusters(struct inode *inode, int cluster)
> >> if (cluster != fatent.entry + 1) {
> >> int nr_clus = fatent.entry - first_cl + 1;
> >>
> >> - sb_issue_discard(sb,
> >> - fat_clus_to_blknr(sbi, first_cl),
> >> - nr_clus * sbi->sec_per_clus,
> >> - GFP_NOFS, 0);
> >> -
> >> + fat_issue_discard(sb, first_cl, nr_clus);
> >> first_cl = cluster;
> >> }
> >> }
> >> @@ -683,3 +691,88 @@ out:
> >> unlock_fat(sbi);
> >> return err;
> >> }
> >> +
> >> +int fat_trim_fs(struct super_block *sb, struct fstrim_range *range)
> >> +{
> >> + struct msdos_sb_info *sbi = MSDOS_SB(sb);
> >> + struct fatent_operations *ops = sbi->fatent_ops;
> >> + struct fat_entry fatent;
> >> + unsigned long reada_blocks, reada_mask, cur_block;
> >> + int err = 0, free, count, entry;
> >> + int start, len, minlen, trimmed;
> >> +
> >> + start = range->start >> sb->s_blocksize_bits;
> >> + start = start / sbi->sec_per_clus;
> >> + len = range->len >> sb->s_blocksize_bits;
> >> + len = len / sbi->sec_per_clus;
> >> + minlen = range->minlen >> sb->s_blocksize_bits;
> >> + minlen = minlen / sbi->sec_per_clus;
> >
> > I should have mention that earlier, but you can also adjust mineln
> > according to the discard_granularity, because extents smaller than that
> > would not be discarded anyway.
> I adjust the minlen granularity with fat cluster unit. and FAT don't
> know what's the effected block size. it's role of user side.
Actually it is not and you should check
request_queue->limits.discard_granularity
see: http://www.spinics.net/lists/linux-ext4/msg23145.html
> >
> >> + trimmed = 0;
> >> + count = 0;
> >> +
> >> + lock_fat(sbi);
> >> + if (sbi->free_clusters != -1 && sbi->free_clus_valid)
> >> + goto out;
> >> +
> >> + reada_blocks = FAT_READA_SIZE >> sb->s_blocksize_bits;
> >> + reada_mask = reada_blocks - 1;
> >> + cur_block = 0;
> >> +
> >> + entry = 0;
> >> + free = 0;
> >> + fatent_init(&fatent);
> >> +
> >> + if (start < FAT_START_ENT)
> >> + start = FAT_START_ENT;
> >
> > You're not using start anywhere.
> >
> >> +
> >> + fatent_set_entry(&fatent, FAT_START_ENT);
> It should be fatent_set_entry(&fatent, start);
> >> +
> >> + while (count < sbi->max_cluster) {
> >> + if (fatent.entry >= sbi->max_cluster)
> >> + fatent.entry = FAT_START_ENT;
> >> + /* readahead of fat blocks */
> >> + if ((cur_block & reada_mask) == 0) {
> >> + unsigned long rest = sbi->fat_length - cur_block;
> >> + fat_ent_reada(sb, &fatent, min(reada_blocks, rest));
> >> + }
> >> + cur_block++;
> >> +
> >> + err = fat_ent_read_block(sb, &fatent);
> >> + if (err)
> >> + goto out;
> >> +
> >> + do {
> >> + if (ops->ent_get(&fatent) == FAT_ENT_FREE) {
> >> + free++;
> >> + if (!entry)
> >> + entry = fatent.entry;
> >> + if (free >= (len - trimmed) && free >= minlen) {
> >
> > It seems to me that you are using len as number of bytes to trim. This
> > is not right and I am sorry for not explaining it correctly. "len" is
> > supposed to be a number of bytes you want to "investigate" from the start.
> > So it means that you will trim every single free extent bigger than minlen
> > between "start" byte and "start + len" byte of the underlying device, or
> > partition.
> No. len is adjusted at fat cluster number. it's not used byte unit.
> I think it's easy to compare with fat internal units.
Does not matter what units are you using for len. I it just that you are
checking for (free >= (len - trimmed)) which is wrong because len is not
meant to be "overall length of trimmed data" but rather "overall length
of data to walk through and check for free extents" see ext4
implementation for reference.
Thanks!
-Lukas
>
> I hope fat peoples comment this one.
>
> Thank you,
> Kyungmin Park
> >
> >> + fat_issue_discard(sb, entry, free);
> >> + trimmed += free;
> >> + }
> >> + if (trimmed >= len)
> >> + goto done;
> >> + } else if (entry) {
> >> + if (free >= minlen) {
> >> + fat_issue_discard(sb, entry, free);
> >> + trimmed += free;
> >> + }
> >> + if (trimmed >= len)
> >> + goto done;
> >> + free = 0;
> >> + entry = 0;
> >> + }
> >> + count++;
> >> + } while (fat_ent_next(sbi, &fatent));
> >> + }
> >> + if (free >= minlen) {
> >> + fat_issue_discard(sb, entry, free);
> >> + trimmed += free;
> >> + }
> >> +done:
> >> + range->len = (trimmed * sbi->sec_per_clus) << sb->s_blocksize_bits;
> >> + fatent_brelse(&fatent);
> >> +out:
> >> + unlock_fat(sbi);
> >> + return err;
> >> +}
> >> diff --git a/fs/fat/file.c b/fs/fat/file.c
> >> index 7257752..05e6545 100644
> >> --- a/fs/fat/file.c
> >> +++ b/fs/fat/file.c
> >> @@ -125,6 +125,34 @@ long fat_generic_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
> >> return fat_ioctl_get_attributes(inode, user_attr);
> >> case FAT_IOCTL_SET_ATTRIBUTES:
> >> return fat_ioctl_set_attributes(filp, user_attr);
> >> + case FITRIM:
> >> + {
> >> + struct super_block *sb = inode->i_sb;
> >> + struct request_queue *q = bdev_get_queue(sb->s_bdev);
> >> + struct fstrim_range range;
> >> + int ret = 0;
> >> +
> >> + if (!capable(CAP_SYS_ADMIN))
> >> + return -EPERM;
> >> +
> >> + if (!blk_queue_discard(q))
> >> + return -EOPNOTSUPP;
> >> +
> >> + if (copy_from_user(&range, (struct fstrim_range *)arg,
> >> + sizeof(range)))
> >> + return -EFAULT;
> >> +
> >> + ret = fat_trim_fs(sb, &range);
> >> + if (ret < 0)
> >> + return ret;
> >> +
> >> + if (copy_to_user((struct fstrim_range *)arg, &range,
> >> + sizeof(range)))
> >> + return -EFAULT;
> >> +
> >> + return 0;
> >> + }
> >> +
> >> default:
> >> return -ENOTTY; /* Inappropriate ioctl for device */
> >> }
> >>
> >
> > --
> >
>
--