Github/littlefs
1
Fork 0
Code Actions Activity

Switched to using mid to tell which files belong in a compaction

Browse Source 
This avoids the previous issues with block state for null inlined data,
and we're already testing the rid anyways for splits.

In theory we don't need the block for inlined data at all, but it is
convenient as it allows us to use the existing internal rbyd/data APIs
without needing to move data around. Though it may be worth looking into
alternative layouts at some point.
This commit is contained in:
Christopher Haster
2023-09-27 15:11:40 -05:00
parent 69993da7e1
commit 0638b09d18
1 changed files with 13 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lfs.c
+13 -14
View File
@@ -5452,7 +5452,7 @@ static int lfsr_mdir_compact__(lfs_t *lfs, lfsr_mdir_t *mdir_,
opened;
opened = opened->next) {
lfsr_file_t *file = (lfsr_file_t*)opened;
if (lfsr_inlined_hasshrub(&file->inlined)
if (lfsr_inlined_hassprout(&file->inlined)
&& file->inlined.u.data.u.disk.block
== data.u.disk.block
&& file->inlined.u.data.u.disk.off
@@ -5550,15 +5550,15 @@ static int lfsr_mdir_compact__(lfs_t *lfs, lfsr_mdir_t *mdir_,
lfsr_file_t *file = (lfsr_file_t*)opened;
// belongs to our mdir?
if (lfsr_file_isunsynced(file)
&& lfsr_inlined_block(&file->inlined) == mdir->u.rbyd.block
&& (file->m.mdir.mid & lfsr_midbmask(lfs))
== (mdir->mid & lfsr_midbmask(lfs))
&& (file->m.mdir.mid & lfsr_midrmask(lfs)) >= start_rid
&& (lfs_size_t)(file->m.mdir.mid & lfsr_midrmask(lfs))
< (lfs_size_t)end_rid) {
// inlined null? we don't write a tag for these but we do need to
// update them, they may have been clobbered in a failed commit
if (lfsr_inlined_hasnull(&file->inlined)) {
file->inlined_.u.data = LFSR_DATA_DISK(
mdir_->u.rbyd.block, 0, 0);
file->inlined_.u.data = LFSR_DATA_DISK(0, 0, 0);
// inlined data?
} else if (lfsr_inlined_hassprout(&file->inlined)) {
@@ -5677,7 +5677,8 @@ static lfs_ssize_t lfsr_mdir_estimate_(lfs_t *lfs, const lfsr_mdir_t *mdir,
lfsr_file_t *file = (lfsr_file_t*)opened;
// belongs to our mdir?
if (lfsr_file_isunsynced(file)
&& lfsr_inlined_block(&file->inlined) == mdir->u.rbyd.block
&& (file->m.mdir.mid & lfsr_midbmask(lfs))
== (mdir->mid & lfsr_midbmask(lfs))
&& (file->m.mdir.mid & lfsr_midrmask(lfs)) == rid) {
// inlined data?
if (lfsr_inlined_hassprout(&file->inlined)) {
@@ -8351,6 +8352,13 @@ int lfsr_file_opencfg(lfs_t *lfs, lfsr_file_t *file,
}
}
// setup file state
file->flags = flags;
file->cfg = cfg;
file->pos = 0;
// default inlined state
file->inlined.u.data = LFSR_DATA_DISK(0, 0, 0);
// lookup our parent
lfsr_tag_t tag;
lfsr_did_t did;
@@ -8363,15 +8371,6 @@ int lfsr_file_opencfg(lfs_t *lfs, lfsr_file_t *file,
return err;
}
// setup file state
file->flags = flags;
file->cfg = cfg;
file->pos = 0;
// default inlined state
// note we still keep track of the block when inlined is null, it's
// a nice invariant to have
file->inlined.u.data = LFSR_DATA_DISK(file->m.mdir.u.m.blocks[0], 0, 0);
// creating a new entry?
if (err == LFS_ERR_NOENT) {
if (!lfsr_flags_iscreat(flags)) {