erofs: implement encoded extent metadata

Implement the extent metadata parsing described in the previous commit.
For 16-byte and 32-byte extent records, currently it is just a trivial
binary search without considering the last access footprint, but it can
be optimized for better sequential performance later.

Tail fragments are supported, but ztailpacking feature is not
for simplicity.

Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Acked-by: Chao Yu <chao@kernel.org>
Link: https://lore.kernel.org/r/20250310095459.2620647-9-hsiangkao@linux.alibaba.com

diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index f26191f..4ac188d 100644 (file)
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -263,7 +263,10 @@ struct erofs_inode {
                        unsigned short z_advise;
                        unsigned char  z_algorithmtype[2];
                        unsigned char  z_lclusterbits;
-                       unsigned long  z_tailextent_headlcn;
+                       union {
+                               u64    z_tailextent_headlcn;
+                               u64    z_extents;
+                       };
                        erofs_off_t    z_fragmentoff;
                        unsigned short z_idata_size;
                };

diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index 25d3fa8..8de50df 100644 (file)
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -391,7 +391,7 @@ static int z_erofs_get_extent_decompressedlen(struct z_erofs_maprecorder *m)
        return 0;
 }
 
-static int z_erofs_do_map_blocks(struct inode *inode,
+static int z_erofs_map_blocks_fo(struct inode *inode,
                                 struct erofs_map_blocks *map, int flags)
 {
        struct erofs_inode *vi = EROFS_I(inode);
@@ -409,6 +409,14 @@ static int z_erofs_do_map_blocks(struct inode *inode,
        unsigned long long ofs, end;
 
        ofs = flags & EROFS_GET_BLOCKS_FINDTAIL ? inode->i_size - 1 : map->m_la;
+       if (fragment && !(flags & EROFS_GET_BLOCKS_FINDTAIL) &&
+           !vi->z_tailextent_headlcn) {
+               map->m_la = 0;
+               map->m_llen = inode->i_size;
+               map->m_flags = EROFS_MAP_MAPPED |
+                       EROFS_MAP_FULL_MAPPED | EROFS_MAP_FRAGMENT;
+               return 0;
+       }
        initial_lcn = ofs >> lclusterbits;
        endoff = ofs & ((1 << lclusterbits) - 1);
 
@@ -526,6 +534,115 @@ unmap_out:
        return err;
 }
 
+static int z_erofs_map_blocks_ext(struct inode *inode,
+                                 struct erofs_map_blocks *map, int flags)
+{
+       struct erofs_inode *vi = EROFS_I(inode);
+       struct super_block *sb = inode->i_sb;
+       bool interlaced = vi->z_advise & Z_EROFS_ADVISE_INTERLACED_PCLUSTER;
+       unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
+       erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
+                                  vi->inode_isize + vi->xattr_isize), recsz);
+       erofs_off_t lend = inode->i_size;
+       erofs_off_t l, r, mid, pa, la, lstart;
+       struct z_erofs_extent *ext;
+       unsigned int fmt;
+       bool last;
+
+       map->m_flags = 0;
+       if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
+               if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
+                       ext = erofs_read_metabuf(&map->buf, sb, pos, true);
+                       if (IS_ERR(ext))
+                               return PTR_ERR(ext);
+                       pa = le64_to_cpu(*(__le64 *)ext);
+                       pos += sizeof(__le64);
+                       lstart = 0;
+               } else {
+                       lstart = map->m_la >> vi->z_lclusterbits;
+                       pa = EROFS_NULL_ADDR;
+               }
+
+               for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
+                       ext = erofs_read_metabuf(&map->buf, sb, pos, true);
+                       if (IS_ERR(ext))
+                               return PTR_ERR(ext);
+                       map->m_plen = le32_to_cpu(ext->plen);
+                       if (pa != EROFS_NULL_ADDR) {
+                               map->m_pa = pa;
+                               pa += map->m_plen & Z_EROFS_EXTENT_PLEN_MASK;
+                       } else {
+                               map->m_pa = le32_to_cpu(ext->pstart_lo);
+                       }
+                       pos += recsz;
+               }
+               last = (lstart >= round_up(lend, 1 << vi->z_lclusterbits));
+               lend = min(lstart, lend);
+               lstart -= 1 << vi->z_lclusterbits;
+       } else {
+               lstart = lend;
+               for (l = 0, r = vi->z_extents; l < r; ) {
+                       mid = l + (r - l) / 2;
+                       ext = erofs_read_metabuf(&map->buf, sb,
+                                                pos + mid * recsz, true);
+                       if (IS_ERR(ext))
+                               return PTR_ERR(ext);
+
+                       la = le32_to_cpu(ext->lstart_lo);
+                       pa = le32_to_cpu(ext->pstart_lo) |
+                               (u64)le32_to_cpu(ext->pstart_hi) << 32;
+                       if (recsz > offsetof(struct z_erofs_extent, lstart_hi))
+                               la |= (u64)le32_to_cpu(ext->lstart_hi) << 32;
+
+                       if (la > map->m_la) {
+                               r = mid;
+                               lend = la;
+                       } else {
+                               l = mid + 1;
+                               if (map->m_la == la)
+                                       r = min(l + 1, r);
+                               lstart = la;
+                               map->m_plen = le32_to_cpu(ext->plen);
+                               map->m_pa = pa;
+                       }
+               }
+               last = (l >= vi->z_extents);
+       }
+
+       if (lstart < lend) {
+               map->m_la = lstart;
+               if (last && (vi->z_advise & Z_EROFS_ADVISE_FRAGMENT_PCLUSTER)) {
+                       map->m_flags |= EROFS_MAP_MAPPED | EROFS_MAP_FRAGMENT;
+                       vi->z_fragmentoff = map->m_plen;
+                       if (recsz >= offsetof(struct z_erofs_extent, pstart_lo))
+                               vi->z_fragmentoff |= map->m_pa << 32;
+               } else if (map->m_plen) {
+                       map->m_flags |= EROFS_MAP_MAPPED |
+                               EROFS_MAP_FULL_MAPPED | EROFS_MAP_ENCODED;
+                       fmt = map->m_plen >> Z_EROFS_EXTENT_PLEN_FMT_BIT;
+                       if (fmt)
+                               map->m_algorithmformat = fmt - 1;
+                       else if (interlaced && !erofs_blkoff(sb, map->m_pa))
+                               map->m_algorithmformat =
+                                       Z_EROFS_COMPRESSION_INTERLACED;
+                       else
+                               map->m_algorithmformat =
+                                       Z_EROFS_COMPRESSION_SHIFTED;
+                       if (map->m_plen & Z_EROFS_EXTENT_PLEN_PARTIAL)
+                               map->m_flags |= EROFS_MAP_PARTIAL_REF;
+                       map->m_plen &= Z_EROFS_EXTENT_PLEN_MASK;
+               }
+       }
+       map->m_llen = lend - map->m_la;
+       if (!last && map->m_llen < sb->s_blocksize) {
+               erofs_err(sb, "extent too small %llu @ offset %llu of nid %llu",
+                         map->m_llen, map->m_la, vi->nid);
+               DBG_BUGON(1);
+               return -EFSCORRUPTED;
+       }
+       return 0;
+}
+
 static int z_erofs_fill_inode_lazy(struct inode *inode)
 {
        struct erofs_inode *const vi = EROFS_I(inode);
@@ -570,6 +687,13 @@ static int z_erofs_fill_inode_lazy(struct inode *inode)
        }
        vi->z_advise = le16_to_cpu(h->h_advise);
        vi->z_lclusterbits = sb->s_blocksize_bits + (h->h_clusterbits & 15);
+       if (vi->datalayout == EROFS_INODE_COMPRESSED_FULL &&
+           (vi->z_advise & Z_EROFS_ADVISE_EXTENTS)) {
+               vi->z_extents = le32_to_cpu(h->h_extents_lo) |
+                       ((u64)le16_to_cpu(h->h_extents_hi) << 32);
+               goto done;
+       }
+
        vi->z_algorithmtype[0] = h->h_algorithmtype & 15;
        vi->z_algorithmtype[1] = h->h_algorithmtype >> 4;
        if (vi->z_advise & Z_EROFS_ADVISE_FRAGMENT_PCLUSTER)
@@ -609,7 +733,7 @@ static int z_erofs_fill_inode_lazy(struct inode *inode)
                        .buf = __EROFS_BUF_INITIALIZER
                };
 
-               err = z_erofs_do_map_blocks(inode, &map,
+               err = z_erofs_map_blocks_fo(inode, &map,
                                            EROFS_GET_BLOCKS_FINDTAIL);
                erofs_put_metabuf(&map.buf);
                if (err < 0)
@@ -640,15 +764,11 @@ int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
        } else {
                err = z_erofs_fill_inode_lazy(inode);
                if (!err) {
-                       if ((vi->z_advise & Z_EROFS_ADVISE_FRAGMENT_PCLUSTER) &&
-                           !vi->z_tailextent_headlcn) {
-                               map->m_la = 0;
-                               map->m_llen = inode->i_size;
-                               map->m_flags = EROFS_MAP_MAPPED |
-                                       EROFS_MAP_FULL_MAPPED | EROFS_MAP_FRAGMENT;
-                       } else {
-                               err = z_erofs_do_map_blocks(inode, map, flags);
-                       }
+                       if (vi->datalayout == EROFS_INODE_COMPRESSED_FULL &&
+                           (vi->z_advise & Z_EROFS_ADVISE_EXTENTS))
+                               err = z_erofs_map_blocks_ext(inode, map, flags);
+                       else
+                               err = z_erofs_map_blocks_fo(inode, map, flags);
                }
                if (!err && (map->m_flags & EROFS_MAP_ENCODED) &&
                    unlikely(map->m_plen > Z_EROFS_PCLUSTER_MAX_SIZE ||