From 4eea22f01bb4fdba1aab4430c33adbe88d9d4985 Mon Sep 17 00:00:00 2001 From: Mandy Kirkconnell Date: Tue, 14 Mar 2006 13:29:52 +1100 Subject: [XFS] 929045 567344 This mod re-organizes some of the in-core file extent code to prepare for an upcoming mod which will introduce multi-level in-core extent allocations. Although the in-core extent management is using a new code path in this mod, the functionality remains the same. Major changes include: - Introduce 10 new subroutines which re-orgainze the existing code but do NOT change functionality: xfs_iext_get_ext() xfs_iext_insert() xfs_iext_add() xfs_iext_remove() xfs_iext_remove_inline() xfs_iext_remove_direct() xfs_iext_realloc_direct() xfs_iext_direct_to_inline() xfs_iext_inline_to_direct() xfs_iext_destroy() - Remove 2 subroutines (functionality moved to new subroutines above): xfs_iext_realloc() -replaced by xfs_iext_add() and xfs_iext_remove() xfs_bmap_insert_exlist() - replaced by xfs_iext_insert() xfs_bmap_delete_exlist() - replaced by xfs_iext_remove() - Replace all hard-coded (indexed) extent assignments with a call to xfs_iext_get_ext() - Replace all extent record pointer arithmetic (ep++, ep--, base + lastx,..) with calls to xfs_iext_get_ext() - Update comments to remove the idea of a single "extent list" and introduce "extent record" terminology instead SGI-PV: 928864 SGI-Modid: xfs-linux-melb:xfs-kern:207390a Signed-off-by: Mandy Kirkconnell Signed-off-by: Nathan Scott --- fs/xfs/xfs_inode.c | 470 +++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 351 insertions(+), 119 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 1d7f5a7e063..6459395a0e4 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -76,16 +76,18 @@ STATIC int xfs_iformat_btree(xfs_inode_t *, xfs_dinode_t *, int); */ STATIC void xfs_validate_extents( - xfs_bmbt_rec_t *ep, + xfs_ifork_t *ifp, int nrecs, int disk, xfs_exntfmt_t fmt) { + xfs_bmbt_rec_t *ep; xfs_bmbt_irec_t irec; xfs_bmbt_rec_t rec; int i; for (i = 0; i < nrecs; i++) { + ep = xfs_iext_get_ext(ifp, i); rec.l0 = get_unaligned((__uint64_t*)&ep->l0); rec.l1 = get_unaligned((__uint64_t*)&ep->l1); if (disk) @@ -94,11 +96,10 @@ xfs_validate_extents( xfs_bmbt_get_all(&rec, &irec); if (fmt == XFS_EXTFMT_NOSTATE) ASSERT(irec.br_state == XFS_EXT_NORM); - ep++; } } #else /* DEBUG */ -#define xfs_validate_extents(ep, nrecs, disk, fmt) +#define xfs_validate_extents(ifp, nrecs, disk, fmt) #endif /* DEBUG */ /* @@ -597,7 +598,6 @@ xfs_iformat_extents( xfs_bmbt_rec_t *ep, *dp; xfs_ifork_t *ifp; int nex; - int real_size; int size; int i; @@ -619,23 +619,20 @@ xfs_iformat_extents( return XFS_ERROR(EFSCORRUPTED); } - real_size = 0; + ifp->if_real_bytes = 0; if (nex == 0) ifp->if_u1.if_extents = NULL; else if (nex <= XFS_INLINE_EXTS) ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; - else { - ifp->if_u1.if_extents = kmem_alloc(size, KM_SLEEP); - ASSERT(ifp->if_u1.if_extents != NULL); - real_size = size; - } + else + xfs_iext_add(ifp, 0, nex); + ifp->if_bytes = size; - ifp->if_real_bytes = real_size; if (size) { dp = (xfs_bmbt_rec_t *) XFS_DFORK_PTR(dip, whichfork); - xfs_validate_extents(dp, nex, 1, XFS_EXTFMT_INODE(ip)); - ep = ifp->if_u1.if_extents; - for (i = 0; i < nex; i++, ep++, dp++) { + xfs_validate_extents(ifp, nex, 1, XFS_EXTFMT_INODE(ip)); + for (i = 0; i < nex; i++, dp++) { + ep = xfs_iext_get_ext(ifp, i); ep->l0 = INT_GET(get_unaligned((__uint64_t*)&dp->l0), ARCH_CONVERT); ep->l1 = INT_GET(get_unaligned((__uint64_t*)&dp->l1), @@ -646,7 +643,7 @@ xfs_iformat_extents( if (whichfork != XFS_DATA_FORK || XFS_EXTFMT_INODE(ip) == XFS_EXTFMT_NOSTATE) if (unlikely(xfs_check_nostate_extents( - ifp->if_u1.if_extents, nex))) { + ifp, 0, nex))) { XFS_ERROR_REPORT("xfs_iformat_extents(2)", XFS_ERRLEVEL_LOW, ip->i_mount); @@ -1015,6 +1012,7 @@ xfs_iread_extents( { int error; xfs_ifork_t *ifp; + xfs_extnum_t nextents; size_t size; if (unlikely(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE)) { @@ -1022,26 +1020,24 @@ xfs_iread_extents( ip->i_mount); return XFS_ERROR(EFSCORRUPTED); } - size = XFS_IFORK_NEXTENTS(ip, whichfork) * (uint)sizeof(xfs_bmbt_rec_t); + nextents = XFS_IFORK_NEXTENTS(ip, whichfork); + size = nextents * sizeof(xfs_bmbt_rec_t); ifp = XFS_IFORK_PTR(ip, whichfork); + /* * We know that the size is valid (it's checked in iformat_btree) */ - ifp->if_u1.if_extents = kmem_alloc(size, KM_SLEEP); - ASSERT(ifp->if_u1.if_extents != NULL); ifp->if_lastex = NULLEXTNUM; - ifp->if_bytes = ifp->if_real_bytes = (int)size; + ifp->if_bytes = ifp->if_real_bytes = 0; ifp->if_flags |= XFS_IFEXTENTS; + xfs_iext_add(ifp, 0, nextents); error = xfs_bmap_read_extents(tp, ip, whichfork); if (error) { - kmem_free(ifp->if_u1.if_extents, size); - ifp->if_u1.if_extents = NULL; - ifp->if_bytes = ifp->if_real_bytes = 0; + xfs_iext_destroy(ifp); ifp->if_flags &= ~XFS_IFEXTENTS; return error; } - xfs_validate_extents((xfs_bmbt_rec_t *)ifp->if_u1.if_extents, - XFS_IFORK_NEXTENTS(ip, whichfork), 0, XFS_EXTFMT_INODE(ip)); + xfs_validate_extents(ifp, nextents, 0, XFS_EXTFMT_INODE(ip)); return 0; } @@ -2475,92 +2471,6 @@ xfs_iroot_realloc( } -/* - * This is called when the amount of space needed for if_extents - * is increased or decreased. The change in size is indicated by - * the number of extents that need to be added or deleted in the - * ext_diff parameter. - * - * If the amount of space needed has decreased below the size of the - * inline buffer, then switch to using the inline buffer. Otherwise, - * use kmem_realloc() or kmem_alloc() to adjust the size of the buffer - * to what is needed. - * - * ip -- the inode whose if_extents area is changing - * ext_diff -- the change in the number of extents, positive or negative, - * requested for the if_extents array. - */ -void -xfs_iext_realloc( - xfs_inode_t *ip, - int ext_diff, - int whichfork) -{ - int byte_diff; - xfs_ifork_t *ifp; - int new_size; - uint rnew_size; - - if (ext_diff == 0) { - return; - } - - ifp = XFS_IFORK_PTR(ip, whichfork); - byte_diff = ext_diff * (uint)sizeof(xfs_bmbt_rec_t); - new_size = (int)ifp->if_bytes + byte_diff; - ASSERT(new_size >= 0); - - if (new_size == 0) { - if (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext) { - ASSERT(ifp->if_real_bytes != 0); - kmem_free(ifp->if_u1.if_extents, ifp->if_real_bytes); - } - ifp->if_u1.if_extents = NULL; - rnew_size = 0; - } else if (new_size <= sizeof(ifp->if_u2.if_inline_ext)) { - /* - * If the valid extents can fit in if_inline_ext, - * copy them from the malloc'd vector and free it. - */ - if (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext) { - /* - * For now, empty files are format EXTENTS, - * so the if_extents pointer is null. - */ - if (ifp->if_u1.if_extents) { - memcpy(ifp->if_u2.if_inline_ext, - ifp->if_u1.if_extents, new_size); - kmem_free(ifp->if_u1.if_extents, - ifp->if_real_bytes); - } - ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; - } - rnew_size = 0; - } else { - rnew_size = new_size; - if ((rnew_size & (rnew_size - 1)) != 0) - rnew_size = xfs_iroundup(rnew_size); - /* - * Stuck with malloc/realloc. - */ - if (ifp->if_u1.if_extents == ifp->if_u2.if_inline_ext) { - ifp->if_u1.if_extents = (xfs_bmbt_rec_t *) - kmem_alloc(rnew_size, KM_SLEEP); - memcpy(ifp->if_u1.if_extents, ifp->if_u2.if_inline_ext, - sizeof(ifp->if_u2.if_inline_ext)); - } else if (rnew_size != ifp->if_real_bytes) { - ifp->if_u1.if_extents = (xfs_bmbt_rec_t *) - kmem_realloc(ifp->if_u1.if_extents, - rnew_size, - ifp->if_real_bytes, - KM_NOFS); - } - } - ifp->if_real_bytes = rnew_size; - ifp->if_bytes = new_size; -} - - /* * This is called when the amount of space needed for if_data * is increased or decreased. The change in size is indicated by @@ -2723,9 +2633,7 @@ xfs_idestroy_fork( (ifp->if_u1.if_extents != NULL) && (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext)) { ASSERT(ifp->if_real_bytes != 0); - kmem_free(ifp->if_u1.if_extents, ifp->if_real_bytes); - ifp->if_u1.if_extents = NULL; - ifp->if_real_bytes = 0; + xfs_iext_destroy(ifp); } ASSERT(ifp->if_u1.if_extents == NULL || ifp->if_u1.if_extents == ifp->if_u2.if_inline_ext); @@ -2902,16 +2810,15 @@ xfs_iextents_copy( * the delayed ones. There must be at least one * non-delayed extent. */ - ep = ifp->if_u1.if_extents; dest_ep = buffer; copied = 0; for (i = 0; i < nrecs; i++) { + ep = xfs_iext_get_ext(ifp, i); start_block = xfs_bmbt_get_startblock(ep); if (ISNULLSTARTBLOCK(start_block)) { /* * It's a delayed allocation extent, so skip it. */ - ep++; continue; } @@ -2921,11 +2828,10 @@ xfs_iextents_copy( put_unaligned(INT_GET(ep->l1, ARCH_CONVERT), (__uint64_t*)&dest_ep->l1); dest_ep++; - ep++; copied++; } ASSERT(copied != 0); - xfs_validate_extents(buffer, copied, 1, XFS_EXTFMT_INODE(ip)); + xfs_validate_extents(ifp, copied, 1, XFS_EXTFMT_INODE(ip)); return (copied * (uint)sizeof(xfs_bmbt_rec_t)); } @@ -2995,8 +2901,10 @@ xfs_iflush_fork( case XFS_DINODE_FMT_EXTENTS: ASSERT((ifp->if_flags & XFS_IFEXTENTS) || !(iip->ili_format.ilf_fields & extflag[whichfork])); - ASSERT((ifp->if_u1.if_extents != NULL) || (ifp->if_bytes == 0)); - ASSERT((ifp->if_u1.if_extents == NULL) || (ifp->if_bytes > 0)); + ASSERT((xfs_iext_get_ext(ifp, 0) != NULL) || + (ifp->if_bytes == 0)); + ASSERT((xfs_iext_get_ext(ifp, 0) == NULL) || + (ifp->if_bytes > 0)); if ((iip->ili_format.ilf_fields & extflag[whichfork]) && (ifp->if_bytes > 0)) { ASSERT(XFS_IFORK_NEXTENTS(ip, whichfork) > 0); @@ -3704,3 +3612,327 @@ xfs_ilock_trace(xfs_inode_t *ip, int lock, unsigned int lockflags, inst_t *ra) NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL); } #endif + +/* + * Return a pointer to the extent record at file index idx. + */ +xfs_bmbt_rec_t * +xfs_iext_get_ext( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx) /* index of target extent */ +{ + ASSERT(idx >= 0); + if (ifp->if_bytes) { + return &ifp->if_u1.if_extents[idx]; + } else { + return NULL; + } +} + +/* + * Insert new item(s) into the extent records for incore inode + * fork 'ifp'. 'count' new items are inserted at index 'idx'. + */ +void +xfs_iext_insert( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* starting index of new items */ + xfs_extnum_t count, /* number of inserted items */ + xfs_bmbt_irec_t *new) /* items to insert */ +{ + xfs_bmbt_rec_t *ep; /* extent record pointer */ + xfs_extnum_t i; /* extent record index */ + + ASSERT(ifp->if_flags & XFS_IFEXTENTS); + xfs_iext_add(ifp, idx, count); + for (i = idx; i < idx + count; i++, new++) { + ep = xfs_iext_get_ext(ifp, i); + xfs_bmbt_set_all(ep, new); + } +} + +/* + * This is called when the amount of space required for incore file + * extents needs to be increased. The ext_diff parameter stores the + * number of new extents being added and the idx parameter contains + * the extent index where the new extents will be added. If the new + * extents are being appended, then we just need to (re)allocate and + * initialize the space. Otherwise, if the new extents are being + * inserted into the middle of the existing entries, a bit more work + * is required to make room for the new extents to be inserted. The + * caller is responsible for filling in the new extent entries upon + * return. + */ +void +xfs_iext_add( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* index to begin adding exts */ + int ext_diff) /* nubmer of extents to add */ +{ + int byte_diff; /* new bytes being added */ + int new_size; /* size of extents after adding */ + xfs_extnum_t nextents; /* number of extents in file */ + + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + ASSERT((idx >= 0) && (idx <= nextents)); + byte_diff = ext_diff * sizeof(xfs_bmbt_rec_t); + new_size = ifp->if_bytes + byte_diff; + /* + * If the new number of extents (nextents + ext_diff) + * fits inside the inode, then continue to use the inline + * extent buffer. + */ + if (nextents + ext_diff <= XFS_INLINE_EXTS) { + if (idx < nextents) { + memmove(&ifp->if_u2.if_inline_ext[idx + ext_diff], + &ifp->if_u2.if_inline_ext[idx], + (nextents - idx) * sizeof(xfs_bmbt_rec_t)); + memset(&ifp->if_u2.if_inline_ext[idx], 0, byte_diff); + } + ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; + ifp->if_real_bytes = 0; + } + /* + * Otherwise use a linear (direct) extent list. + * If the extents are currently inside the inode, + * xfs_iext_realloc_direct will switch us from + * inline to direct extent allocation mode. + */ + else { + xfs_iext_realloc_direct(ifp, new_size); + if (idx < nextents) { + memmove(&ifp->if_u1.if_extents[idx + ext_diff], + &ifp->if_u1.if_extents[idx], + (nextents - idx) * sizeof(xfs_bmbt_rec_t)); + memset(&ifp->if_u1.if_extents[idx], 0, byte_diff); + } + } + ifp->if_bytes = new_size; +} + +/* + * This is called when the amount of space required for incore file + * extents needs to be decreased. The ext_diff parameter stores the + * number of extents to be removed and the idx parameter contains + * the extent index where the extents will be removed from. + */ +void +xfs_iext_remove( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* index to begin removing exts */ + int ext_diff) /* number of extents to remove */ +{ + xfs_extnum_t nextents; /* number of extents in file */ + int new_size; /* size of extents after removal */ + + ASSERT(ext_diff > 0); + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + new_size = (nextents - ext_diff) * sizeof(xfs_bmbt_rec_t); + + if (new_size == 0) { + xfs_iext_destroy(ifp); + } else if (ifp->if_real_bytes) { + xfs_iext_remove_direct(ifp, idx, ext_diff); + } else { + xfs_iext_remove_inline(ifp, idx, ext_diff); + } + ifp->if_bytes = new_size; +} + +/* + * This removes ext_diff extents from the inline buffer, beginning + * at extent index idx. + */ +void +xfs_iext_remove_inline( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* index to begin removing exts */ + int ext_diff) /* number of extents to remove */ +{ + int nextents; /* number of extents in file */ + + ASSERT(idx < XFS_INLINE_EXTS); + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + ASSERT(((nextents - ext_diff) > 0) && + (nextents - ext_diff) < XFS_INLINE_EXTS); + + if (idx + ext_diff < nextents) { + memmove(&ifp->if_u2.if_inline_ext[idx], + &ifp->if_u2.if_inline_ext[idx + ext_diff], + (nextents - (idx + ext_diff)) * + sizeof(xfs_bmbt_rec_t)); + memset(&ifp->if_u2.if_inline_ext[nextents - ext_diff], + 0, ext_diff * sizeof(xfs_bmbt_rec_t)); + } else { + memset(&ifp->if_u2.if_inline_ext[idx], 0, + ext_diff * sizeof(xfs_bmbt_rec_t)); + } +} + +/* + * This removes ext_diff extents from a linear (direct) extent list, + * beginning at extent index idx. If the extents are being removed + * from the end of the list (ie. truncate) then we just need to re- + * allocate the list to remove the extra space. Otherwise, if the + * extents are being removed from the middle of the existing extent + * entries, then we first need to move the extent records beginning + * at idx + ext_diff up in the list to overwrite the records being + * removed, then remove the extra space via kmem_realloc. + */ +void +xfs_iext_remove_direct( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* index to begin removing exts */ + int ext_diff) /* number of extents to remove */ +{ + xfs_extnum_t nextents; /* number of extents in file */ + int new_size; /* size of extents after removal */ + + new_size = ifp->if_bytes - + (ext_diff * sizeof(xfs_bmbt_rec_t)); + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + + if (new_size == 0) { + xfs_iext_destroy(ifp); + return; + } + /* Move extents up in the list (if needed) */ + if (idx + ext_diff < nextents) { + memmove(&ifp->if_u1.if_extents[idx], + &ifp->if_u1.if_extents[idx + ext_diff], + (nextents - (idx + ext_diff)) * + sizeof(xfs_bmbt_rec_t)); + } + memset(&ifp->if_u1.if_extents[nextents - ext_diff], + 0, ext_diff * sizeof(xfs_bmbt_rec_t)); + /* + * Reallocate the direct extent list. If the extents + * will fit inside the inode then xfs_iext_realloc_direct + * will switch from direct to inline extent allocation + * mode for us. + */ + xfs_iext_realloc_direct(ifp, new_size); + ifp->if_bytes = new_size; +} + +/* + * Create, destroy, or resize a linear (direct) block of extents. + */ +void +xfs_iext_realloc_direct( + xfs_ifork_t *ifp, /* inode fork pointer */ + int new_size) /* new size of extents */ +{ + int rnew_size; /* real new size of extents */ + + rnew_size = new_size; + + /* Free extent records */ + if (new_size == 0) { + xfs_iext_destroy(ifp); + } + /* Resize direct extent list and zero any new bytes */ + else if (ifp->if_real_bytes) { + /* Check if extents will fit inside the inode */ + if (new_size <= XFS_INLINE_EXTS * sizeof(xfs_bmbt_rec_t)) { + xfs_iext_direct_to_inline(ifp, new_size / + (uint)sizeof(xfs_bmbt_rec_t)); + ifp->if_bytes = new_size; + return; + } + if ((new_size & (new_size - 1)) != 0) { + rnew_size = xfs_iroundup(new_size); + } + if (rnew_size != ifp->if_real_bytes) { + ifp->if_u1.if_extents = (xfs_bmbt_rec_t *) + kmem_realloc(ifp->if_u1.if_extents, + rnew_size, + ifp->if_real_bytes, + KM_SLEEP); + } + if (rnew_size > ifp->if_real_bytes) { + memset(&ifp->if_u1.if_extents[ifp->if_bytes / + (uint)sizeof(xfs_bmbt_rec_t)], 0, + rnew_size - ifp->if_real_bytes); + } + } + /* + * Switch from the inline extent buffer to a direct + * extent list. Be sure to include the inline extent + * bytes in new_size. + */ + else { + new_size += ifp->if_bytes; + if ((new_size & (new_size - 1)) != 0) { + rnew_size = xfs_iroundup(new_size); + } + xfs_iext_inline_to_direct(ifp, rnew_size); + } + ifp->if_real_bytes = rnew_size; + ifp->if_bytes = new_size; +} + +/* + * Switch from linear (direct) extent records to inline buffer. + */ +void +xfs_iext_direct_to_inline( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t nextents) /* number of extents in file */ +{ + ASSERT(ifp->if_flags & XFS_IFEXTENTS); + ASSERT(nextents <= XFS_INLINE_EXTS); + /* + * The inline buffer was zeroed when we switched + * from inline to direct extent allocation mode, + * so we don't need to clear it here. + */ + memcpy(ifp->if_u2.if_inline_ext, ifp->if_u1.if_extents, + nextents * sizeof(xfs_bmbt_rec_t)); + kmem_free(ifp->if_u1.if_extents, KM_SLEEP); + ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; + ifp->if_real_bytes = 0; +} + +/* + * Switch from inline buffer to linear (direct) extent records. + * new_size should already be rounded up to the next power of 2 + * by the caller (when appropriate), so use new_size as it is. + * However, since new_size may be rounded up, we can't update + * if_bytes here. It is the caller's responsibility to update + * if_bytes upon return. + */ +void +xfs_iext_inline_to_direct( + xfs_ifork_t *ifp, /* inode fork pointer */ + int new_size) /* number of extents in file */ +{ + ifp->if_u1.if_extents = (xfs_bmbt_rec_t *) + kmem_alloc(new_size, KM_SLEEP); + memset(ifp->if_u1.if_extents, 0, new_size); + if (ifp->if_bytes) { + memcpy(ifp->if_u1.if_extents, ifp->if_u2.if_inline_ext, + ifp->if_bytes); + memset(ifp->if_u2.if_inline_ext, 0, XFS_INLINE_EXTS * + sizeof(xfs_bmbt_rec_t)); + } + ifp->if_real_bytes = new_size; +} + +/* + * Free incore file extents. + */ +void +xfs_iext_destroy( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + if (ifp->if_real_bytes) { + kmem_free(ifp->if_u1.if_extents, ifp->if_real_bytes); + } else if (ifp->if_bytes) { + memset(ifp->if_u2.if_inline_ext, 0, XFS_INLINE_EXTS * + sizeof(xfs_bmbt_rec_t)); + } + ifp->if_u1.if_extents = NULL; + ifp->if_real_bytes = 0; + ifp->if_bytes = 0; +} -- cgit v1.2.3-70-g09d2 From 0293ce3a9fd1b34c933a96577a8ba737b681cf75 Mon Sep 17 00:00:00 2001 From: Mandy Kirkconnell Date: Tue, 14 Mar 2006 13:30:23 +1100 Subject: [XFS] 929045 567344 This mod introduces multi-level in-core file extent functionality, building upon the new layout introduced in mod xfs-linux:xfs-kern:207390a. The new multi-level extent allocations are only required for heavily fragmented files, so the old-style linear extent list is used on files until the extents reach a pre-determined size of 4k. 4k buffers are used because this is the system page size on Linux i386 and systems with larger page sizes don't seem to gain much, if anything, by using their native page size as the extent buffer size. Also, using 4k extent buffers everywhere provides a consistent interface for CXFS across different platforms. The 4k extent buffers are managed by an indirection array (xfs_ext_irec_t) which is basically just a pointer array with a bit of extra information to keep track of the number of extents in each buffer as well as the extent offset of each buffer. Major changes include: - Add multi-level in-core file extent functionality to the xfs_iext_ subroutines introduced in mod: xfs-linux:xfs-kern:207390a - Introduce 13 new subroutines which add functionality for multi-level in-core file extents: xfs_iext_add_indirect_multi() xfs_iext_remove_indirect() xfs_iext_realloc_indirect() xfs_iext_indirect_to_direct() xfs_iext_bno_to_irec() xfs_iext_idx_to_irec() xfs_iext_irec_init() xfs_iext_irec_new() xfs_iext_irec_remove() xfs_iext_irec_compact() xfs_iext_irec_compact_pages() xfs_iext_irec_compact_full() xfs_iext_irec_update_extoffs() SGI-PV: 928864 SGI-Modid: xfs-linux-melb:xfs-kern:207393a Signed-off-by: Mandy Kirkconnell Signed-off-by: Nathan Scott --- fs/xfs/xfs_bmap.c | 59 +++- fs/xfs/xfs_bmap.h | 18 ++ fs/xfs/xfs_bmap_btree.h | 8 - fs/xfs/xfs_inode.c | 715 +++++++++++++++++++++++++++++++++++++++++++++++- fs/xfs/xfs_inode.h | 52 +++- 5 files changed, 826 insertions(+), 26 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c index 53c47a181f8..81a95b684b6 100644 --- a/fs/xfs/xfs_bmap.c +++ b/fs/xfs/xfs_bmap.c @@ -3423,6 +3423,7 @@ xfs_bmap_local_to_extents( xfs_bmap_forkoff_reset(args.mp, ip, whichfork); xfs_idata_realloc(ip, -ifp->if_bytes, whichfork); xfs_iext_add(ifp, 0, 1); + ASSERT((ifp->if_flags & (XFS_IFEXTENTS|XFS_IFEXTIREC)) == XFS_IFEXTENTS); ep = xfs_iext_get_ext(ifp, 0); xfs_bmbt_set_allf(ep, 0, args.fsbno, 1, XFS_EXT_NORM); xfs_bmap_trace_post_update(fname, "new", ip, 0, whichfork); @@ -3551,6 +3552,54 @@ xfs_bmap_do_search_extents( return ep; } +/* + * Call xfs_bmap_do_search_extents() to search for the extent + * record containing block bno. If in multi-level in-core extent + * allocation mode, find and extract the target extent buffer, + * otherwise just use the direct extent list. + */ +xfs_bmbt_rec_t * /* pointer to found extent entry */ +xfs_bmap_search_multi_extents( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_fileoff_t bno, /* block number searched for */ + int *eofp, /* out: end of file found */ + xfs_extnum_t *lastxp, /* out: last extent index */ + xfs_bmbt_irec_t *gotp, /* out: extent entry found */ + xfs_bmbt_irec_t *prevp) /* out: previous extent entry found */ +{ + xfs_bmbt_rec_t *base; /* base of extent records */ + xfs_bmbt_rec_t *ep; /* extent record pointer */ + xfs_ext_irec_t *erp = NULL; /* indirection array pointer */ + xfs_extnum_t lastx; /* last extent index */ + xfs_extnum_t nextents; /* number of file extents */ + + /* + * For multi-level extent allocation mode, find the + * target extent list and pass only the contiguous + * list to xfs_bmap_do_search_extents. Convert lastx + * from a file extent index to an index within the + * target extent list. + */ + if (ifp->if_flags & XFS_IFEXTIREC) { + int erp_idx = 0; + erp = xfs_iext_bno_to_irec(ifp, bno, &erp_idx); + base = erp->er_extbuf; + nextents = erp->er_extcount; + lastx = ifp->if_lastex - erp->er_extoff; + } else { + base = &ifp->if_u1.if_extents[0]; + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + lastx = ifp->if_lastex; + } + ep = xfs_bmap_do_search_extents(base, lastx, nextents, bno, + eofp, lastxp, gotp, prevp); + /* Convert lastx back to file-based index */ + if (ifp->if_flags & XFS_IFEXTIREC) { + *lastxp += erp->er_extoff; + } + return ep; +} + /* * Search the extents list for the inode, for the extent containing bno. * If bno lies in a hole, point to the next entry. If bno lies past eof, @@ -3569,20 +3618,14 @@ xfs_bmap_search_extents( xfs_bmbt_irec_t *prevp) /* out: previous extent entry found */ { xfs_ifork_t *ifp; /* inode fork pointer */ - xfs_bmbt_rec_t *base; /* base of extent list */ - xfs_extnum_t lastx; /* last extent index used */ - xfs_extnum_t nextents; /* number of file extents */ xfs_bmbt_rec_t *ep; /* extent record pointer */ int rt; /* realtime flag */ XFS_STATS_INC(xs_look_exlist); ifp = XFS_IFORK_PTR(ip, whichfork); - lastx = ifp->if_lastex; - nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); - base = &ifp->if_u1.if_extents[0]; - ep = xfs_bmap_do_search_extents(base, lastx, nextents, bno, eofp, - lastxp, gotp, prevp); + ep = xfs_bmap_search_multi_extents(ifp, bno, eofp, lastxp, gotp, prevp); + rt = (whichfork == XFS_DATA_FORK) && XFS_IS_REALTIME_INODE(ip); if (unlikely(!rt && !gotp->br_startblock && (*lastxp != NULLEXTNUM))) { cmn_err(CE_PANIC,"Access to block zero: fs: <%s> inode: %lld " diff --git a/fs/xfs/xfs_bmap.h b/fs/xfs/xfs_bmap.h index 4c05f95452c..011ccaa9a1c 100644 --- a/fs/xfs/xfs_bmap.h +++ b/fs/xfs/xfs_bmap.h @@ -352,6 +352,24 @@ xfs_check_nostate_extents( xfs_extnum_t idx, xfs_extnum_t num); +/* + * Call xfs_bmap_do_search_extents() to search for the extent + * record containing block bno. If in multi-level in-core extent + * allocation mode, find and extract the target extent buffer, + * otherwise just use the direct extent list. + */ +xfs_bmbt_rec_t * +xfs_bmap_search_multi_extents(struct xfs_ifork *, xfs_fileoff_t, int *, + xfs_extnum_t *, xfs_bmbt_irec_t *, xfs_bmbt_irec_t *); + +/* + * Search an extent list for the extent which includes block + * bno. + */ +xfs_bmbt_rec_t *xfs_bmap_do_search_extents(xfs_bmbt_rec_t *, + xfs_extnum_t, xfs_extnum_t, xfs_fileoff_t, int *, + xfs_extnum_t *, xfs_bmbt_irec_t *, xfs_bmbt_irec_t *); + #endif /* __KERNEL__ */ #endif /* __XFS_BMAP_H__ */ diff --git a/fs/xfs/xfs_bmap_btree.h b/fs/xfs/xfs_bmap_btree.h index e095a2d344a..6478cfa0e53 100644 --- a/fs/xfs/xfs_bmap_btree.h +++ b/fs/xfs/xfs_bmap_btree.h @@ -372,14 +372,6 @@ extern int xfs_bmbt_get_rec(struct xfs_btree_cur *, xfs_fileoff_t *, xfs_exntst_t *, int *); #endif -/* - * Search an extent list for the extent which includes block - * bno. - */ -xfs_bmbt_rec_t *xfs_bmap_do_search_extents(xfs_bmbt_rec_t *, - xfs_extnum_t, xfs_extnum_t, xfs_fileoff_t, int *, - xfs_extnum_t *, xfs_bmbt_irec_t *, xfs_bmbt_irec_t *); - #endif /* __KERNEL__ */ #endif /* __XFS_BMAP_BTREE_H__ */ diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 6459395a0e4..580fa075803 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2630,8 +2630,9 @@ xfs_idestroy_fork( ifp->if_real_bytes = 0; } } else if ((ifp->if_flags & XFS_IFEXTENTS) && - (ifp->if_u1.if_extents != NULL) && - (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext)) { + ((ifp->if_flags & XFS_IFEXTIREC) || + ((ifp->if_u1.if_extents != NULL) && + (ifp->if_u1.if_extents != ifp->if_u2.if_inline_ext)))) { ASSERT(ifp->if_real_bytes != 0); xfs_iext_destroy(ifp); } @@ -3622,7 +3623,16 @@ xfs_iext_get_ext( xfs_extnum_t idx) /* index of target extent */ { ASSERT(idx >= 0); - if (ifp->if_bytes) { + if ((ifp->if_flags & XFS_IFEXTIREC) && (idx == 0)) { + return ifp->if_u1.if_ext_irec->er_extbuf; + } else if (ifp->if_flags & XFS_IFEXTIREC) { + xfs_ext_irec_t *erp; /* irec pointer */ + int erp_idx = 0; /* irec index */ + xfs_extnum_t page_idx = idx; /* ext index in target list */ + + erp = xfs_iext_idx_to_irec(ifp, &page_idx, &erp_idx, 0); + return &erp->er_extbuf[page_idx]; + } else if (ifp->if_bytes) { return &ifp->if_u1.if_extents[idx]; } else { return NULL; @@ -3691,6 +3701,7 @@ xfs_iext_add( } ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; ifp->if_real_bytes = 0; + ifp->if_lastex = nextents + ext_diff; } /* * Otherwise use a linear (direct) extent list. @@ -3698,7 +3709,7 @@ xfs_iext_add( * xfs_iext_realloc_direct will switch us from * inline to direct extent allocation mode. */ - else { + else if (nextents + ext_diff <= XFS_LINEAR_EXTS) { xfs_iext_realloc_direct(ifp, new_size); if (idx < nextents) { memmove(&ifp->if_u1.if_extents[idx + ext_diff], @@ -3707,14 +3718,182 @@ xfs_iext_add( memset(&ifp->if_u1.if_extents[idx], 0, byte_diff); } } + /* Indirection array */ + else { + xfs_ext_irec_t *erp; + int erp_idx = 0; + int page_idx = idx; + + ASSERT(nextents + ext_diff > XFS_LINEAR_EXTS); + if (ifp->if_flags & XFS_IFEXTIREC) { + erp = xfs_iext_idx_to_irec(ifp, &page_idx, &erp_idx, 1); + } else { + xfs_iext_irec_init(ifp); + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + erp = ifp->if_u1.if_ext_irec; + } + /* Extents fit in target extent page */ + if (erp && erp->er_extcount + ext_diff <= XFS_LINEAR_EXTS) { + if (page_idx < erp->er_extcount) { + memmove(&erp->er_extbuf[page_idx + ext_diff], + &erp->er_extbuf[page_idx], + (erp->er_extcount - page_idx) * + sizeof(xfs_bmbt_rec_t)); + memset(&erp->er_extbuf[page_idx], 0, byte_diff); + } + erp->er_extcount += ext_diff; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff); + } + /* Insert a new extent page */ + else if (erp) { + xfs_iext_add_indirect_multi(ifp, + erp_idx, page_idx, ext_diff); + } + /* + * If extent(s) are being appended to the last page in + * the indirection array and the new extent(s) don't fit + * in the page, then erp is NULL and erp_idx is set to + * the next index needed in the indirection array. + */ + else { + int count = ext_diff; + + while (count) { + erp = xfs_iext_irec_new(ifp, erp_idx); + erp->er_extcount = count; + count -= MIN(count, (int)XFS_LINEAR_EXTS); + if (count) { + erp_idx++; + } + } + } + } ifp->if_bytes = new_size; } +/* + * This is called when incore extents are being added to the indirection + * array and the new extents do not fit in the target extent list. The + * erp_idx parameter contains the irec index for the target extent list + * in the indirection array, and the idx parameter contains the extent + * index within the list. The number of extents being added is stored + * in the count parameter. + * + * |-------| |-------| + * | | | | idx - number of extents before idx + * | idx | | count | + * | | | | count - number of extents being inserted at idx + * |-------| |-------| + * | count | | nex2 | nex2 - number of extents after idx + count + * |-------| |-------| + */ +void +xfs_iext_add_indirect_multi( + xfs_ifork_t *ifp, /* inode fork pointer */ + int erp_idx, /* target extent irec index */ + xfs_extnum_t idx, /* index within target list */ + int count) /* new extents being added */ +{ + int byte_diff; /* new bytes being added */ + xfs_ext_irec_t *erp; /* pointer to irec entry */ + xfs_extnum_t ext_diff; /* number of extents to add */ + xfs_extnum_t ext_cnt; /* new extents still needed */ + xfs_extnum_t nex2; /* extents after idx + count */ + xfs_bmbt_rec_t *nex2_ep = NULL; /* temp list for nex2 extents */ + int nlists; /* number of irec's (lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + nex2 = erp->er_extcount - idx; + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + + /* + * Save second part of target extent list + * (all extents past */ + if (nex2) { + byte_diff = nex2 * sizeof(xfs_bmbt_rec_t); + nex2_ep = (xfs_bmbt_rec_t *) kmem_alloc(byte_diff, KM_SLEEP); + memmove(nex2_ep, &erp->er_extbuf[idx], byte_diff); + erp->er_extcount -= nex2; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, -nex2); + memset(&erp->er_extbuf[idx], 0, byte_diff); + } + + /* + * Add the new extents to the end of the target + * list, then allocate new irec record(s) and + * extent buffer(s) as needed to store the rest + * of the new extents. + */ + ext_cnt = count; + ext_diff = MIN(ext_cnt, (int)XFS_LINEAR_EXTS - erp->er_extcount); + if (ext_diff) { + erp->er_extcount += ext_diff; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff); + ext_cnt -= ext_diff; + } + while (ext_cnt) { + erp_idx++; + erp = xfs_iext_irec_new(ifp, erp_idx); + ext_diff = MIN(ext_cnt, (int)XFS_LINEAR_EXTS); + erp->er_extcount = ext_diff; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, ext_diff); + ext_cnt -= ext_diff; + } + + /* Add nex2 extents back to indirection array */ + if (nex2) { + xfs_extnum_t ext_avail; + int i; + + byte_diff = nex2 * sizeof(xfs_bmbt_rec_t); + ext_avail = XFS_LINEAR_EXTS - erp->er_extcount; + i = 0; + /* + * If nex2 extents fit in the current page, append + * nex2_ep after the new extents. + */ + if (nex2 <= ext_avail) { + i = erp->er_extcount; + } + /* + * Otherwise, check if space is available in the + * next page. + */ + else if ((erp_idx < nlists - 1) && + (nex2 <= (ext_avail = XFS_LINEAR_EXTS - + ifp->if_u1.if_ext_irec[erp_idx+1].er_extcount))) { + erp_idx++; + erp++; + /* Create a hole for nex2 extents */ + memmove(&erp->er_extbuf[nex2], erp->er_extbuf, + erp->er_extcount * sizeof(xfs_bmbt_rec_t)); + } + /* + * Final choice, create a new extent page for + * nex2 extents. + */ + else { + erp_idx++; + erp = xfs_iext_irec_new(ifp, erp_idx); + } + memmove(&erp->er_extbuf[i], nex2_ep, byte_diff); + kmem_free(nex2_ep, byte_diff); + erp->er_extcount += nex2; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, nex2); + } +} + /* * This is called when the amount of space required for incore file * extents needs to be decreased. The ext_diff parameter stores the * number of extents to be removed and the idx parameter contains * the extent index where the extents will be removed from. + * + * If the amount of space needed has decreased below the linear + * limit, XFS_IEXT_BUFSZ, then switch to using the contiguous + * extent array. Otherwise, use kmem_realloc() to adjust the + * size to what is needed. */ void xfs_iext_remove( @@ -3731,6 +3910,8 @@ xfs_iext_remove( if (new_size == 0) { xfs_iext_destroy(ifp); + } else if (ifp->if_flags & XFS_IFEXTIREC) { + xfs_iext_remove_indirect(ifp, idx, ext_diff); } else if (ifp->if_real_bytes) { xfs_iext_remove_direct(ifp, idx, ext_diff); } else { @@ -3751,6 +3932,7 @@ xfs_iext_remove_inline( { int nextents; /* number of extents in file */ + ASSERT(!(ifp->if_flags & XFS_IFEXTIREC)); ASSERT(idx < XFS_INLINE_EXTS); nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); ASSERT(((nextents - ext_diff) > 0) && @@ -3788,6 +3970,7 @@ xfs_iext_remove_direct( xfs_extnum_t nextents; /* number of extents in file */ int new_size; /* size of extents after removal */ + ASSERT(!(ifp->if_flags & XFS_IFEXTIREC)); new_size = ifp->if_bytes - (ext_diff * sizeof(xfs_bmbt_rec_t)); nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); @@ -3815,6 +3998,84 @@ xfs_iext_remove_direct( ifp->if_bytes = new_size; } +/* + * This is called when incore extents are being removed from the + * indirection array and the extents being removed span multiple extent + * buffers. The idx parameter contains the file extent index where we + * want to begin removing extents, and the count parameter contains + * how many extents need to be removed. + * + * |-------| |-------| + * | nex1 | | | nex1 - number of extents before idx + * |-------| | count | + * | | | | count - number of extents being removed at idx + * | count | |-------| + * | | | nex2 | nex2 - number of extents after idx + count + * |-------| |-------| + */ +void +xfs_iext_remove_indirect( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t idx, /* index to begin removing extents */ + int count) /* number of extents to remove */ +{ + xfs_ext_irec_t *erp; /* indirection array pointer */ + int erp_idx = 0; /* indirection array index */ + xfs_extnum_t ext_cnt; /* extents left to remove */ + xfs_extnum_t ext_diff; /* extents to remove in current list */ + xfs_extnum_t nex1; /* number of extents before idx */ + xfs_extnum_t nex2; /* extents after idx + count */ + int nlists; /* entries in indirecton array */ + int page_idx = idx; /* index in target extent list */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + erp = xfs_iext_idx_to_irec(ifp, &page_idx, &erp_idx, 0); + ASSERT(erp != NULL); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + nex1 = page_idx; + ext_cnt = count; + while (ext_cnt) { + nex2 = MAX((erp->er_extcount - (nex1 + ext_cnt)), 0); + ext_diff = MIN(ext_cnt, (erp->er_extcount - nex1)); + /* + * Check for deletion of entire list; + * xfs_iext_irec_remove() updates extent offsets. + */ + if (ext_diff == erp->er_extcount) { + xfs_iext_irec_remove(ifp, erp_idx); + ext_cnt -= ext_diff; + nex1 = 0; + if (ext_cnt) { + ASSERT(erp_idx < ifp->if_real_bytes / + XFS_IEXT_BUFSZ); + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + nex1 = 0; + continue; + } else { + break; + } + } + /* Move extents up (if needed) */ + if (nex2) { + memmove(&erp->er_extbuf[nex1], + &erp->er_extbuf[nex1 + ext_diff], + nex2 * sizeof(xfs_bmbt_rec_t)); + } + /* Zero out rest of page */ + memset(&erp->er_extbuf[nex1 + nex2], 0, (XFS_IEXT_BUFSZ - + ((nex1 + nex2) * sizeof(xfs_bmbt_rec_t)))); + /* Update remaining counters */ + erp->er_extcount -= ext_diff; + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, -ext_diff); + ext_cnt -= ext_diff; + nex1 = 0; + erp_idx++; + erp++; + } + ifp->if_bytes -= count * sizeof(xfs_bmbt_rec_t); + xfs_iext_irec_compact(ifp); +} + /* * Create, destroy, or resize a linear (direct) block of extents. */ @@ -3827,6 +4088,10 @@ xfs_iext_realloc_direct( rnew_size = new_size; + ASSERT(!(ifp->if_flags & XFS_IFEXTIREC) || + ((new_size >= 0) && (new_size <= XFS_IEXT_BUFSZ) && + (new_size != ifp->if_real_bytes))); + /* Free extent records */ if (new_size == 0) { xfs_iext_destroy(ifp); @@ -3919,6 +4184,60 @@ xfs_iext_inline_to_direct( ifp->if_real_bytes = new_size; } +/* + * Resize an extent indirection array to new_size bytes. + */ +void +xfs_iext_realloc_indirect( + xfs_ifork_t *ifp, /* inode fork pointer */ + int new_size) /* new indirection array size */ +{ + int nlists; /* number of irec's (ex lists) */ + int size; /* current indirection array size */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + size = nlists * sizeof(xfs_ext_irec_t); + ASSERT(ifp->if_real_bytes); + ASSERT((new_size >= 0) && (new_size != size)); + if (new_size == 0) { + xfs_iext_destroy(ifp); + } else { + ifp->if_u1.if_ext_irec = (xfs_ext_irec_t *) + kmem_realloc(ifp->if_u1.if_ext_irec, + new_size, size, KM_SLEEP); + } +} + +/* + * Switch from indirection array to linear (direct) extent allocations. + */ +void +xfs_iext_indirect_to_direct( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + xfs_bmbt_rec_t *ep; /* extent record pointer */ + xfs_extnum_t nextents; /* number of extents in file */ + int size; /* size of file extents */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + ASSERT(nextents <= XFS_LINEAR_EXTS); + size = nextents * sizeof(xfs_bmbt_rec_t); + + xfs_iext_irec_compact_full(ifp); + ASSERT(ifp->if_real_bytes == XFS_IEXT_BUFSZ); + + ep = ifp->if_u1.if_ext_irec->er_extbuf; + kmem_free(ifp->if_u1.if_ext_irec, sizeof(xfs_ext_irec_t)); + ifp->if_flags &= ~XFS_IFEXTIREC; + ifp->if_u1.if_extents = ep; + ifp->if_bytes = size; + if (nextents < XFS_LINEAR_EXTS) { + xfs_iext_realloc_direct(ifp, size); + } +} + /* * Free incore file extents. */ @@ -3926,7 +4245,16 @@ void xfs_iext_destroy( xfs_ifork_t *ifp) /* inode fork pointer */ { - if (ifp->if_real_bytes) { + if (ifp->if_flags & XFS_IFEXTIREC) { + int erp_idx; + int nlists; + + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + for (erp_idx = nlists - 1; erp_idx >= 0 ; erp_idx--) { + xfs_iext_irec_remove(ifp, erp_idx); + } + ifp->if_flags &= ~XFS_IFEXTIREC; + } else if (ifp->if_real_bytes) { kmem_free(ifp->if_u1.if_extents, ifp->if_real_bytes); } else if (ifp->if_bytes) { memset(ifp->if_u2.if_inline_ext, 0, XFS_INLINE_EXTS * @@ -3936,3 +4264,380 @@ xfs_iext_destroy( ifp->if_real_bytes = 0; ifp->if_bytes = 0; } + +/* + * Return a pointer to the indirection array entry containing the + * extent record for filesystem block bno. Store the index of the + * target irec in *erp_idxp. + */ +xfs_ext_irec_t * +xfs_iext_bno_to_irec( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_fileoff_t bno, /* block number to search for */ + int *erp_idxp) /* irec index of target ext list */ +{ + xfs_ext_irec_t *erp = NULL; /* indirection array pointer */ + xfs_ext_irec_t *erp_next; /* next indirection array entry */ + xfs_extnum_t erp_idx; /* indirection array index */ + int nlists; /* number of extent irec's (lists) */ + int high; /* binary search upper limit */ + int low; /* binary search lower limit */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + erp_idx = 0; + low = 0; + high = nlists - 1; + while (low <= high) { + erp_idx = (low + high) >> 1; + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + erp_next = erp_idx < nlists - 1 ? erp + 1 : NULL; + if (bno < xfs_bmbt_get_startoff(erp->er_extbuf)) { + high = erp_idx - 1; + } else if (erp_next && bno >= + xfs_bmbt_get_startoff(erp_next->er_extbuf)) { + low = erp_idx + 1; + } else { + break; + } + } + *erp_idxp = erp_idx; + return erp; +} + +/* + * Return a pointer to the indirection array entry containing the + * extent record at file extent index *idxp. Store the index of the + * target irec in *erp_idxp and store the page index of the target + * extent record in *idxp. + */ +xfs_ext_irec_t * +xfs_iext_idx_to_irec( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_extnum_t *idxp, /* extent index (file -> page) */ + int *erp_idxp, /* pointer to target irec */ + int realloc) /* new bytes were just added */ +{ + xfs_ext_irec_t *prev; /* pointer to previous irec */ + xfs_ext_irec_t *erp = NULL; /* pointer to current irec */ + int erp_idx; /* indirection array index */ + int nlists; /* number of irec's (ex lists) */ + int high; /* binary search upper limit */ + int low; /* binary search lower limit */ + xfs_extnum_t page_idx = *idxp; /* extent index in target list */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + ASSERT(page_idx >= 0 && page_idx <= + ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t)); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + erp_idx = 0; + low = 0; + high = nlists - 1; + + /* Binary search extent irec's */ + while (low <= high) { + erp_idx = (low + high) >> 1; + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + prev = erp_idx > 0 ? erp - 1 : NULL; + if (page_idx < erp->er_extoff || (page_idx == erp->er_extoff && + realloc && prev && prev->er_extcount < XFS_LINEAR_EXTS)) { + high = erp_idx - 1; + } else if (page_idx > erp->er_extoff + erp->er_extcount || + (page_idx == erp->er_extoff + erp->er_extcount && + !realloc)) { + low = erp_idx + 1; + } else if (page_idx == erp->er_extoff + erp->er_extcount && + erp->er_extcount == XFS_LINEAR_EXTS) { + ASSERT(realloc); + page_idx = 0; + erp_idx++; + erp = erp_idx < nlists ? erp + 1 : NULL; + break; + } else { + page_idx -= erp->er_extoff; + break; + } + } + *idxp = page_idx; + *erp_idxp = erp_idx; + return(erp); +} + +/* + * Allocate and initialize an indirection array once the space needed + * for incore extents increases above XFS_IEXT_BUFSZ. + */ +void +xfs_iext_irec_init( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + xfs_ext_irec_t *erp; /* indirection array pointer */ + xfs_extnum_t nextents; /* number of extents in file */ + + ASSERT(!(ifp->if_flags & XFS_IFEXTIREC)); + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + ASSERT(nextents <= XFS_LINEAR_EXTS); + + erp = (xfs_ext_irec_t *) + kmem_alloc(sizeof(xfs_ext_irec_t), KM_SLEEP); + + if (nextents == 0) { + ifp->if_u1.if_extents = (xfs_bmbt_rec_t *) + kmem_alloc(XFS_IEXT_BUFSZ, KM_SLEEP); + } else if (!ifp->if_real_bytes) { + xfs_iext_inline_to_direct(ifp, XFS_IEXT_BUFSZ); + } else if (ifp->if_real_bytes < XFS_IEXT_BUFSZ) { + xfs_iext_realloc_direct(ifp, XFS_IEXT_BUFSZ); + } + erp->er_extbuf = ifp->if_u1.if_extents; + erp->er_extcount = nextents; + erp->er_extoff = 0; + + ifp->if_flags |= XFS_IFEXTIREC; + ifp->if_real_bytes = XFS_IEXT_BUFSZ; + ifp->if_bytes = nextents * sizeof(xfs_bmbt_rec_t); + ifp->if_u1.if_ext_irec = erp; + + return; +} + +/* + * Allocate and initialize a new entry in the indirection array. + */ +xfs_ext_irec_t * +xfs_iext_irec_new( + xfs_ifork_t *ifp, /* inode fork pointer */ + int erp_idx) /* index for new irec */ +{ + xfs_ext_irec_t *erp; /* indirection array pointer */ + int i; /* loop counter */ + int nlists; /* number of irec's (ex lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + + /* Resize indirection array */ + xfs_iext_realloc_indirect(ifp, ++nlists * + sizeof(xfs_ext_irec_t)); + /* + * Move records down in the array so the + * new page can use erp_idx. + */ + erp = ifp->if_u1.if_ext_irec; + for (i = nlists - 1; i > erp_idx; i--) { + memmove(&erp[i], &erp[i-1], sizeof(xfs_ext_irec_t)); + } + ASSERT(i == erp_idx); + + /* Initialize new extent record */ + erp = ifp->if_u1.if_ext_irec; + erp[erp_idx].er_extbuf = (xfs_bmbt_rec_t *) + kmem_alloc(XFS_IEXT_BUFSZ, KM_SLEEP); + ifp->if_real_bytes = nlists * XFS_IEXT_BUFSZ; + memset(erp[erp_idx].er_extbuf, 0, XFS_IEXT_BUFSZ); + erp[erp_idx].er_extcount = 0; + erp[erp_idx].er_extoff = erp_idx > 0 ? + erp[erp_idx-1].er_extoff + erp[erp_idx-1].er_extcount : 0; + return (&erp[erp_idx]); +} + +/* + * Remove a record from the indirection array. + */ +void +xfs_iext_irec_remove( + xfs_ifork_t *ifp, /* inode fork pointer */ + int erp_idx) /* irec index to remove */ +{ + xfs_ext_irec_t *erp; /* indirection array pointer */ + int i; /* loop counter */ + int nlists; /* number of irec's (ex lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + if (erp->er_extbuf) { + xfs_iext_irec_update_extoffs(ifp, erp_idx + 1, + -erp->er_extcount); + kmem_free(erp->er_extbuf, XFS_IEXT_BUFSZ); + } + /* Compact extent records */ + erp = ifp->if_u1.if_ext_irec; + for (i = erp_idx; i < nlists - 1; i++) { + memmove(&erp[i], &erp[i+1], sizeof(xfs_ext_irec_t)); + } + /* + * Manually free the last extent record from the indirection + * array. A call to xfs_iext_realloc_indirect() with a size + * of zero would result in a call to xfs_iext_destroy() which + * would in turn call this function again, creating a nasty + * infinite loop. + */ + if (--nlists) { + xfs_iext_realloc_indirect(ifp, + nlists * sizeof(xfs_ext_irec_t)); + } else { + kmem_free(ifp->if_u1.if_ext_irec, + sizeof(xfs_ext_irec_t)); + } + ifp->if_real_bytes = nlists * XFS_IEXT_BUFSZ; +} + +/* + * This is called to clean up large amounts of unused memory allocated + * by the indirection array. Before compacting anything though, verify + * that the indirection array is still needed and switch back to the + * linear extent list (or even the inline buffer) if possible. The + * compaction policy is as follows: + * + * Full Compaction: Extents fit into a single page (or inline buffer) + * Full Compaction: Extents occupy less than 10% of allocated space + * Partial Compaction: Extents occupy > 10% and < 50% of allocated space + * No Compaction: Extents occupy at least 50% of allocated space + */ +void +xfs_iext_irec_compact( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + xfs_extnum_t nextents; /* number of extents in file */ + int nlists; /* number of irec's (ex lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + + if (nextents == 0) { + xfs_iext_destroy(ifp); + } else if (nextents <= XFS_INLINE_EXTS) { + xfs_iext_indirect_to_direct(ifp); + xfs_iext_direct_to_inline(ifp, nextents); + } else if (nextents <= XFS_LINEAR_EXTS) { + xfs_iext_indirect_to_direct(ifp); + } else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 3) { + xfs_iext_irec_compact_full(ifp); + } else if (nextents < (nlists * XFS_LINEAR_EXTS) >> 1) { + xfs_iext_irec_compact_pages(ifp); + } +} + +/* + * Combine extents from neighboring extent pages. + */ +void +xfs_iext_irec_compact_pages( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + xfs_ext_irec_t *erp, *erp_next;/* pointers to irec entries */ + int erp_idx = 0; /* indirection array index */ + int nlists; /* number of irec's (ex lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + while (erp_idx < nlists - 1) { + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + erp_next = erp + 1; + if (erp_next->er_extcount <= + (XFS_LINEAR_EXTS - erp->er_extcount)) { + memmove(&erp->er_extbuf[erp->er_extcount], + erp_next->er_extbuf, erp_next->er_extcount * + sizeof(xfs_bmbt_rec_t)); + erp->er_extcount += erp_next->er_extcount; + /* + * Free page before removing extent record + * so er_extoffs don't get modified in + * xfs_iext_irec_remove. + */ + kmem_free(erp_next->er_extbuf, XFS_IEXT_BUFSZ); + erp_next->er_extbuf = NULL; + xfs_iext_irec_remove(ifp, erp_idx + 1); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + } else { + erp_idx++; + } + } +} + +/* + * Fully compact the extent records managed by the indirection array. + */ +void +xfs_iext_irec_compact_full( + xfs_ifork_t *ifp) /* inode fork pointer */ +{ + xfs_bmbt_rec_t *ep, *ep_next; /* extent record pointers */ + xfs_ext_irec_t *erp, *erp_next; /* extent irec pointers */ + int erp_idx = 0; /* extent irec index */ + int ext_avail; /* empty entries in ex list */ + int ext_diff; /* number of exts to add */ + int nlists; /* number of irec's (ex lists) */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + erp = ifp->if_u1.if_ext_irec; + ep = &erp->er_extbuf[erp->er_extcount]; + erp_next = erp + 1; + ep_next = erp_next->er_extbuf; + while (erp_idx < nlists - 1) { + ext_avail = XFS_LINEAR_EXTS - erp->er_extcount; + ext_diff = MIN(ext_avail, erp_next->er_extcount); + memcpy(ep, ep_next, ext_diff * sizeof(xfs_bmbt_rec_t)); + erp->er_extcount += ext_diff; + erp_next->er_extcount -= ext_diff; + /* Remove next page */ + if (erp_next->er_extcount == 0) { + /* + * Free page before removing extent record + * so er_extoffs don't get modified in + * xfs_iext_irec_remove. + */ + kmem_free(erp_next->er_extbuf, + erp_next->er_extcount * sizeof(xfs_bmbt_rec_t)); + erp_next->er_extbuf = NULL; + xfs_iext_irec_remove(ifp, erp_idx + 1); + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + /* Update next page */ + } else { + /* Move rest of page up to become next new page */ + memmove(erp_next->er_extbuf, ep_next, + erp_next->er_extcount * sizeof(xfs_bmbt_rec_t)); + ep_next = erp_next->er_extbuf; + memset(&ep_next[erp_next->er_extcount], 0, + (XFS_LINEAR_EXTS - erp_next->er_extcount) * + sizeof(xfs_bmbt_rec_t)); + } + if (erp->er_extcount == XFS_LINEAR_EXTS) { + erp_idx++; + if (erp_idx < nlists) + erp = &ifp->if_u1.if_ext_irec[erp_idx]; + else + break; + } + ep = &erp->er_extbuf[erp->er_extcount]; + erp_next = erp + 1; + ep_next = erp_next->er_extbuf; + } +} + +/* + * This is called to update the er_extoff field in the indirection + * array when extents have been added or removed from one of the + * extent lists. erp_idx contains the irec index to begin updating + * at and ext_diff contains the number of extents that were added + * or removed. + */ +void +xfs_iext_irec_update_extoffs( + xfs_ifork_t *ifp, /* inode fork pointer */ + int erp_idx, /* irec index to update */ + int ext_diff) /* number of new extents */ +{ + int i; /* loop counter */ + int nlists; /* number of irec's (ex lists */ + + ASSERT(ifp->if_flags & XFS_IFEXTIREC); + nlists = ifp->if_real_bytes / XFS_IEXT_BUFSZ; + for (i = erp_idx; i < nlists; i++) { + ifp->if_u1.if_ext_irec[i].er_extoff += ext_diff; + } +} diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index 740b73fabd2..3c1df1d642f 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -24,11 +24,38 @@ #define XFS_DATA_FORK 0 #define XFS_ATTR_FORK 1 +/* + * The following xfs_ext_irec_t struct introduces a second (top) level + * to the in-core extent allocation scheme. These structs are allocated + * in a contiguous block, creating an indirection array where each entry + * (irec) contains a pointer to a buffer of in-core extent records which + * it manages. Each extent buffer is 4k in size, since 4k is the system + * page size on Linux i386 and systems with larger page sizes don't seem + * to gain much, if anything, by using their native page size as the + * extent buffer size. Also, using 4k extent buffers everywhere provides + * a consistent interface for CXFS across different platforms. + * + * There is currently no limit on the number of irec's (extent lists) + * allowed, so heavily fragmented files may require an indirection array + * which spans multiple system pages of memory. The number of extents + * which would require this amount of contiguous memory is very large + * and should not cause problems in the foreseeable future. However, + * if the memory needed for the contiguous array ever becomes a problem, + * it is possible that a third level of indirection may be required. + */ +typedef struct xfs_ext_irec { + xfs_bmbt_rec_t *er_extbuf; /* block of extent records */ + xfs_extnum_t er_extoff; /* extent offset in file */ + xfs_extnum_t er_extcount; /* number of extents in page/block */ +} xfs_ext_irec_t; + /* * File incore extent information, present for each of data & attr forks. */ -#define XFS_INLINE_EXTS 2 -#define XFS_INLINE_DATA 32 +#define XFS_IEXT_BUFSZ 4096 +#define XFS_LINEAR_EXTS (XFS_IEXT_BUFSZ / (uint)sizeof(xfs_bmbt_rec_t)) +#define XFS_INLINE_EXTS 2 +#define XFS_INLINE_DATA 32 typedef struct xfs_ifork { int if_bytes; /* bytes in if_u1 */ int if_real_bytes; /* bytes allocated in if_u1 */ @@ -39,6 +66,7 @@ typedef struct xfs_ifork { xfs_extnum_t if_lastex; /* last if_extents used */ union { xfs_bmbt_rec_t *if_extents; /* linear map file exts */ + xfs_ext_irec_t *if_ext_irec; /* irec map file exts */ char *if_data; /* inline file data */ } if_u1; union { @@ -61,9 +89,10 @@ typedef struct xfs_ifork { /* * Per-fork incore inode flags. */ -#define XFS_IFINLINE 0x0001 /* Inline data is read in */ -#define XFS_IFEXTENTS 0x0002 /* All extent pointers are read in */ -#define XFS_IFBROOT 0x0004 /* i_broot points to the bmap b-tree root */ +#define XFS_IFINLINE 0x01 /* Inline data is read in */ +#define XFS_IFEXTENTS 0x02 /* All extent pointers are read in */ +#define XFS_IFBROOT 0x04 /* i_broot points to the bmap b-tree root */ +#define XFS_IFEXTIREC 0x08 /* Indirection array of extent blocks */ /* * Flags for xfs_imap() and xfs_dilocate(). @@ -438,13 +467,26 @@ xfs_bmbt_rec_t *xfs_iext_get_ext(xfs_ifork_t *, xfs_extnum_t); void xfs_iext_insert(xfs_ifork_t *, xfs_extnum_t, xfs_extnum_t, xfs_bmbt_irec_t *); void xfs_iext_add(xfs_ifork_t *, xfs_extnum_t, int); +void xfs_iext_add_indirect_multi(xfs_ifork_t *, int, xfs_extnum_t, int); void xfs_iext_remove(xfs_ifork_t *, xfs_extnum_t, int); void xfs_iext_remove_inline(xfs_ifork_t *, xfs_extnum_t, int); void xfs_iext_remove_direct(xfs_ifork_t *, xfs_extnum_t, int); +void xfs_iext_remove_indirect(xfs_ifork_t *, xfs_extnum_t, int); void xfs_iext_realloc_direct(xfs_ifork_t *, int); +void xfs_iext_realloc_indirect(xfs_ifork_t *, int); +void xfs_iext_indirect_to_direct(xfs_ifork_t *); void xfs_iext_direct_to_inline(xfs_ifork_t *, xfs_extnum_t); void xfs_iext_inline_to_direct(xfs_ifork_t *, int); void xfs_iext_destroy(xfs_ifork_t *); +xfs_ext_irec_t *xfs_iext_bno_to_irec(xfs_ifork_t *, xfs_fileoff_t, int *); +xfs_ext_irec_t *xfs_iext_idx_to_irec(xfs_ifork_t *, xfs_extnum_t *, int *, int); +void xfs_iext_irec_init(xfs_ifork_t *); +xfs_ext_irec_t *xfs_iext_irec_new(xfs_ifork_t *, int); +void xfs_iext_irec_remove(xfs_ifork_t *, int); +void xfs_iext_irec_compact(xfs_ifork_t *); +void xfs_iext_irec_compact_pages(xfs_ifork_t *); +void xfs_iext_irec_compact_full(xfs_ifork_t *); +void xfs_iext_irec_update_extoffs(xfs_ifork_t *, int, int); #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount)) -- cgit v1.2.3-70-g09d2 From 8867bc9bf0aed7181aa72c7c938c6ce830b75166 Mon Sep 17 00:00:00 2001 From: Mandy Kirkconnell Date: Fri, 17 Mar 2006 17:25:04 +1100 Subject: [XFS] There are a few problems with the new xfs_bmap_search_multi_extents() wrapper function that I introduced in mod xfs-linux:xfs-kern:207393a. The function was added as a wrapper around xfs_bmap_do_search_extents() to avoid breaking the top-of-tree CXFS interface. The idea of the function was basically to extract the target extent buffer (if muli- level extent allocation mode), then call xfs_bmap_do_search_extents() with either a pointer to the first extent in the target buffer or a pointer to the first extent in the file, depending on which extent mode was being used. However, in addition to locating the target extent record for block bno, xfs_bmap_do_search_extents() also sets four parameters needed by the caller: *lastx, *eofp, *gotp, *prevp. Passing only the target extent buffer to xfs_bmap_do_search_extents() causes *eofp to be set incorrectly if the extent is at the end of the target list but there are actually more extents in the next er_extbuf. Likewise, if the extent is the first one in the buffer but NOT the first in the file, *prevp is incorrectly set to NULL. Adding the needed functionality to xfs_bmap_search_multi_extents() to re-set any incorrectly set fields is redundant and makes the call to xfs_bmap_do_search_extents() not make much sense when multi-level extent allocation mode is being used. This mod basically extracts the two functional components from xfs_bmap_do_search_extents(), with the intent of obsoleting/removing xfs_bmap_do_search_extents() after the CXFS mult-level in-core extent changes are checked in. The two components are: 1) The binary search to locate the target extent record, and 2) Setting the four parameters needed by the caller (*lastx, *eofp, *gotp, *prevp). Component 1: I created a new function in xfs_inode.c called xfs_iext_bno_to_ext(), which executes the binary search to find the target extent record. xfs_bmap_search_multi_extents() has been modified to call xfs_iext_bno_to_ext() rather than xfs_bmap_do_search_extents(). Component 2: The parameter setting functionality has been added to xfs_bmap_search_multi_extents(), eliminating the need for xfs_bmap_do_search_extents(). These changes make the removal of xfs_bmap_do_search_extents() trival once the CXFS changes are in place. They also allow us to maintain the current XFS interface, using the new search function introduced in mod xfs-linux:xfs-kern:207393a. SGI-PV: 928864 SGI-Modid: xfs-linux-melb:xfs-kern:207866a Signed-off-by: Mandy Kirkconnell Signed-off-by: Nathan Scott --- fs/xfs/xfs_bmap.c | 57 ++++++++++++++++++++++-------------------- fs/xfs/xfs_inode.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++++++++-- fs/xfs/xfs_inode.h | 1 + 3 files changed, 102 insertions(+), 29 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c index 64a02eaf1df..2d702e4a74a 100644 --- a/fs/xfs/xfs_bmap.c +++ b/fs/xfs/xfs_bmap.c @@ -3575,10 +3575,11 @@ xfs_bmap_do_search_extents( } /* - * Call xfs_bmap_do_search_extents() to search for the extent - * record containing block bno. If in multi-level in-core extent - * allocation mode, find and extract the target extent buffer, - * otherwise just use the direct extent list. + * Search the extent records for the entry containing block bno. + * If bno lies in a hole, point to the next entry. If bno lies + * past eof, *eofp will be set, and *prevp will contain the last + * entry (null if none). Else, *lastxp will be set to the index + * of the found entry; *gotp will contain the entry. */ xfs_bmbt_rec_t * /* pointer to found extent entry */ xfs_bmap_search_multi_extents( @@ -3589,36 +3590,38 @@ xfs_bmap_search_multi_extents( xfs_bmbt_irec_t *gotp, /* out: extent entry found */ xfs_bmbt_irec_t *prevp) /* out: previous extent entry found */ { - xfs_bmbt_rec_t *base; /* base of extent records */ xfs_bmbt_rec_t *ep; /* extent record pointer */ - xfs_ext_irec_t *erp = NULL; /* indirection array pointer */ xfs_extnum_t lastx; /* last extent index */ - xfs_extnum_t nextents; /* number of file extents */ /* - * For multi-level extent allocation mode, find the - * target extent list and pass only the contiguous - * list to xfs_bmap_do_search_extents. Convert lastx - * from a file extent index to an index within the - * target extent list. + * Initialize the extent entry structure to catch access to + * uninitialized br_startblock field. */ - if (ifp->if_flags & XFS_IFEXTIREC) { - int erp_idx = 0; - erp = xfs_iext_bno_to_irec(ifp, bno, &erp_idx); - base = erp->er_extbuf; - nextents = erp->er_extcount; - lastx = ifp->if_lastex - erp->er_extoff; - } else { - base = &ifp->if_u1.if_extents[0]; - nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); - lastx = ifp->if_lastex; + gotp->br_startoff = 0xffa5a5a5a5a5a5a5LL; + gotp->br_blockcount = 0xa55a5a5a5a5a5a5aLL; + gotp->br_state = XFS_EXT_INVALID; +#if XFS_BIG_BLKNOS + gotp->br_startblock = 0xffffa5a5a5a5a5a5LL; +#else + gotp->br_startblock = 0xffffa5a5; +#endif + prevp->br_startoff = NULLFILEOFF; + + ep = xfs_iext_bno_to_ext(ifp, bno, &lastx); + if (lastx > 0) { + xfs_bmbt_get_all(xfs_iext_get_ext(ifp, lastx - 1), prevp); } - ep = xfs_bmap_do_search_extents(base, lastx, nextents, bno, - eofp, lastxp, gotp, prevp); - /* Convert lastx back to file-based index */ - if (ifp->if_flags & XFS_IFEXTIREC) { - *lastxp += erp->er_extoff; + if (lastx < (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t))) { + xfs_bmbt_get_all(ep, gotp); + *eofp = 0; + } else { + if (lastx > 0) { + *gotp = *prevp; + } + *eofp = 1; + ep = NULL; } + *lastxp = lastx; return ep; } diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 580fa075803..a16df2d435f 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -4265,12 +4265,81 @@ xfs_iext_destroy( ifp->if_bytes = 0; } +/* + * Return a pointer to the extent record for file system block bno. + */ +xfs_bmbt_rec_t * /* pointer to found extent record */ +xfs_iext_bno_to_ext( + xfs_ifork_t *ifp, /* inode fork pointer */ + xfs_fileoff_t bno, /* block number to search for */ + xfs_extnum_t *idxp) /* index of target extent */ +{ + xfs_bmbt_rec_t *base; /* pointer to first extent */ + xfs_filblks_t blockcount = 0; /* number of blocks in extent */ + xfs_bmbt_rec_t *ep = NULL; /* pointer to target extent */ + xfs_ext_irec_t *erp = NULL; /* indirection array pointer */ + int high; /* upper boundry in search */ + xfs_extnum_t idx = 0; /* index of target extent */ + int low; /* lower boundry in search */ + xfs_extnum_t nextents; /* number of file extents */ + xfs_fileoff_t startoff = 0; /* start offset of extent */ + + nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t); + if (nextents == 0) { + *idxp = 0; + return NULL; + } + low = 0; + if (ifp->if_flags & XFS_IFEXTIREC) { + /* Find target extent list */ + int erp_idx = 0; + erp = xfs_iext_bno_to_irec(ifp, bno, &erp_idx); + base = erp->er_extbuf; + high = erp->er_extcount - 1; + } else { + base = ifp->if_u1.if_extents; + high = nextents - 1; + } + /* Binary search extent records */ + while (low <= high) { + idx = (low + high) >> 1; + ep = base + idx; + startoff = xfs_bmbt_get_startoff(ep); + blockcount = xfs_bmbt_get_blockcount(ep); + if (bno < startoff) { + high = idx - 1; + } else if (bno >= startoff + blockcount) { + low = idx + 1; + } else { + /* Convert back to file-based extent index */ + if (ifp->if_flags & XFS_IFEXTIREC) { + idx += erp->er_extoff; + } + *idxp = idx; + return ep; + } + } + /* Convert back to file-based extent index */ + if (ifp->if_flags & XFS_IFEXTIREC) { + idx += erp->er_extoff; + } + if (bno >= startoff + blockcount) { + if (++idx == nextents) { + ep = NULL; + } else { + ep = xfs_iext_get_ext(ifp, idx); + } + } + *idxp = idx; + return ep; +} + /* * Return a pointer to the indirection array entry containing the * extent record for filesystem block bno. Store the index of the * target irec in *erp_idxp. */ -xfs_ext_irec_t * +xfs_ext_irec_t * /* pointer to found extent record */ xfs_iext_bno_to_irec( xfs_ifork_t *ifp, /* inode fork pointer */ xfs_fileoff_t bno, /* block number to search for */ @@ -4278,7 +4347,7 @@ xfs_iext_bno_to_irec( { xfs_ext_irec_t *erp = NULL; /* indirection array pointer */ xfs_ext_irec_t *erp_next; /* next indirection array entry */ - xfs_extnum_t erp_idx; /* indirection array index */ + int erp_idx; /* indirection array index */ int nlists; /* number of extent irec's (lists) */ int high; /* binary search upper limit */ int low; /* binary search lower limit */ diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index 3c1df1d642f..006396764cb 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -478,6 +478,7 @@ void xfs_iext_indirect_to_direct(xfs_ifork_t *); void xfs_iext_direct_to_inline(xfs_ifork_t *, xfs_extnum_t); void xfs_iext_inline_to_direct(xfs_ifork_t *, int); void xfs_iext_destroy(xfs_ifork_t *); +xfs_bmbt_rec_t *xfs_iext_bno_to_ext(xfs_ifork_t *, xfs_fileoff_t, int *); xfs_ext_irec_t *xfs_iext_bno_to_irec(xfs_ifork_t *, xfs_fileoff_t, int *); xfs_ext_irec_t *xfs_iext_idx_to_irec(xfs_ifork_t *, xfs_extnum_t *, int *, int); void xfs_iext_irec_init(xfs_ifork_t *); -- cgit v1.2.3-70-g09d2 From ec86dc02fdc062d0d298814b1e78b482ab38caf7 Mon Sep 17 00:00:00 2001 From: Nathan Scott Date: Fri, 17 Mar 2006 17:25:36 +1100 Subject: [XFS] Complete transition away from linvfs naming convention, finally. SGI-PV: 947038 SGI-Modid: xfs-linux-melb:xfs-kern:25474a Signed-off-by: Nathan Scott --- fs/xfs/linux-2.6/xfs_aops.c | 12 +++++----- fs/xfs/linux-2.6/xfs_export.c | 10 ++++---- fs/xfs/linux-2.6/xfs_file.c | 30 ++++++++++++------------ fs/xfs/linux-2.6/xfs_fs_subr.c | 6 ++--- fs/xfs/linux-2.6/xfs_ioctl.c | 8 +++---- fs/xfs/linux-2.6/xfs_ioctl32.c | 2 +- fs/xfs/linux-2.6/xfs_iops.c | 48 +++++++++++++++++++------------------- fs/xfs/linux-2.6/xfs_linux.h | 2 +- fs/xfs/linux-2.6/xfs_lrw.c | 2 +- fs/xfs/linux-2.6/xfs_super.c | 52 +++++++++++++++++++----------------------- fs/xfs/linux-2.6/xfs_super.h | 5 ---- fs/xfs/linux-2.6/xfs_vfs.c | 17 +++++++++++++- fs/xfs/linux-2.6/xfs_vfs.h | 3 ++- fs/xfs/linux-2.6/xfs_vnode.c | 6 ++--- fs/xfs/linux-2.6/xfs_vnode.h | 32 +++++++++++++++----------- fs/xfs/xfs_attr.c | 4 ++-- fs/xfs/xfs_dfrag.c | 4 ++-- fs/xfs/xfs_iget.c | 6 ++--- fs/xfs/xfs_inode.c | 4 ++-- fs/xfs/xfs_vnodeops.c | 4 ++-- 20 files changed, 135 insertions(+), 122 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c index 2b610c7ba32..448912f0e75 100644 --- a/fs/xfs/linux-2.6/xfs_aops.c +++ b/fs/xfs/linux-2.6/xfs_aops.c @@ -76,7 +76,7 @@ xfs_page_trace( int mask) { xfs_inode_t *ip; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); loff_t isize = i_size_read(inode); loff_t offset = page_offset(page); int delalloc = -1, unmapped = -1, unwritten = -1; @@ -214,7 +214,7 @@ xfs_alloc_ioend( ioend->io_uptodate = 1; /* cleared if any I/O fails */ ioend->io_list = NULL; ioend->io_type = type; - ioend->io_vnode = LINVFS_GET_VP(inode); + ioend->io_vnode = vn_from_inode(inode); ioend->io_buffer_head = NULL; ioend->io_buffer_tail = NULL; atomic_inc(&ioend->io_vnode->v_iocount); @@ -239,7 +239,7 @@ xfs_map_blocks( xfs_iomap_t *mapp, int flags) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error, nmaps = 1; VOP_BMAP(vp, offset, count, flags, mapp, &nmaps, error); @@ -1229,7 +1229,7 @@ __xfs_get_block( int direct, bmapi_flags_t flags) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); xfs_iomap_t iomap; xfs_off_t offset; ssize_t size; @@ -1371,7 +1371,7 @@ xfs_vm_direct_IO( { struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); xfs_iomap_t iomap; int maps = 1; int error; @@ -1410,7 +1410,7 @@ xfs_vm_bmap( sector_t block) { struct inode *inode = (struct inode *)mapping->host; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error; vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); diff --git a/fs/xfs/linux-2.6/xfs_export.c b/fs/xfs/linux-2.6/xfs_export.c index 391c1353cd4..b768ea910bb 100644 --- a/fs/xfs/linux-2.6/xfs_export.c +++ b/fs/xfs/linux-2.6/xfs_export.c @@ -97,7 +97,7 @@ xfs_fs_encode_fh( int len; int is64 = 0; #if XFS_BIG_INUMS - vfs_t *vfs = LINVFS_GET_VFS(inode->i_sb); + vfs_t *vfs = vfs_from_sb(inode->i_sb); if (!(vfs->vfs_flag & VFS_32BITINODES)) { /* filesystem may contain 64bit inode numbers */ @@ -139,14 +139,14 @@ xfs_fs_get_dentry( vnode_t *vp; struct inode *inode; struct dentry *result; - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); int error; VFS_VGET(vfsp, &vp, (fid_t *)data, error); if (error || vp == NULL) return ERR_PTR(-ESTALE) ; - inode = LINVFS_GET_IP(vp); + inode = vn_to_inode(vp); result = d_alloc_anon(inode); if (!result) { iput(inode); @@ -164,12 +164,12 @@ xfs_fs_get_parent( struct dentry *parent; cvp = NULL; - vp = LINVFS_GET_VP(child->d_inode); + vp = vn_from_inode(child->d_inode); VOP_LOOKUP(vp, &dotdot, &cvp, 0, NULL, NULL, error); if (unlikely(error)) return ERR_PTR(-error); - parent = d_alloc_anon(LINVFS_GET_IP(cvp)); + parent = d_alloc_anon(vn_to_inode(cvp)); if (unlikely(!parent)) { VN_RELE(cvp); return ERR_PTR(-ENOMEM); diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c index c271c993649..185567a6a56 100644 --- a/fs/xfs/linux-2.6/xfs_file.c +++ b/fs/xfs/linux-2.6/xfs_file.c @@ -58,7 +58,7 @@ __xfs_file_read( { struct iovec iov = {buf, count}; struct file *file = iocb->ki_filp; - vnode_t *vp = LINVFS_GET_VP(file->f_dentry->d_inode); + vnode_t *vp = vn_from_inode(file->f_dentry->d_inode); ssize_t rval; BUG_ON(iocb->ki_pos != pos); @@ -102,7 +102,7 @@ __xfs_file_write( struct iovec iov = {(void __user *)buf, count}; struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); ssize_t rval; BUG_ON(iocb->ki_pos != pos); @@ -144,7 +144,7 @@ __xfs_file_readv( loff_t *ppos) { struct inode *inode = file->f_mapping->host; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); struct kiocb kiocb; ssize_t rval; @@ -189,7 +189,7 @@ __xfs_file_writev( loff_t *ppos) { struct inode *inode = file->f_mapping->host; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); struct kiocb kiocb; ssize_t rval; @@ -233,7 +233,7 @@ xfs_file_sendfile( read_actor_t actor, void *target) { - vnode_t *vp = LINVFS_GET_VP(filp->f_dentry->d_inode); + vnode_t *vp = vn_from_inode(filp->f_dentry->d_inode); ssize_t rval; VOP_SENDFILE(vp, filp, ppos, 0, count, actor, target, NULL, rval); @@ -246,7 +246,7 @@ xfs_file_open( struct inode *inode, struct file *filp) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error; if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) @@ -263,7 +263,7 @@ xfs_file_release( struct inode *inode, struct file *filp) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error = 0; if (vp) @@ -279,7 +279,7 @@ xfs_file_fsync( int datasync) { struct inode *inode = dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error; int flags = FSYNC_WAIT; @@ -307,7 +307,7 @@ xfs_vm_nopage( int *type) { struct inode *inode = area->vm_file->f_dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); xfs_mount_t *mp = XFS_VFSTOM(vp->v_vfsp); int error; @@ -340,7 +340,7 @@ xfs_file_readdir( xfs_off_t start_offset, curr_offset; xfs_dirent_t *dbp = NULL; - vp = LINVFS_GET_VP(filp->f_dentry->d_inode); + vp = vn_from_inode(filp->f_dentry->d_inode); ASSERT(vp); /* Try fairly hard to get memory */ @@ -409,7 +409,7 @@ xfs_file_mmap( struct vm_area_struct *vma) { struct inode *ip = filp->f_dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(ip); + vnode_t *vp = vn_from_inode(ip); vattr_t vattr; int error; @@ -437,7 +437,7 @@ xfs_file_ioctl( { int error; struct inode *inode = filp->f_dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); VOP_IOCTL(vp, inode, filp, 0, cmd, (void __user *)arg, error); VMODIFY(vp); @@ -459,7 +459,7 @@ xfs_file_ioctl_invis( { int error; struct inode *inode = filp->f_dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); ASSERT(vp); VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error); @@ -481,7 +481,7 @@ xfs_vm_mprotect( struct vm_area_struct *vma, unsigned int newflags) { - vnode_t *vp = LINVFS_GET_VP(vma->vm_file->f_dentry->d_inode); + vnode_t *vp = vn_from_inode(vma->vm_file->f_dentry->d_inode); int error = 0; if (vp->v_vfsp->vfs_flag & VFS_DMI) { @@ -507,7 +507,7 @@ STATIC int xfs_file_open_exec( struct inode *inode) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); xfs_mount_t *mp = XFS_VFSTOM(vp->v_vfsp); int error = 0; xfs_inode_t *ip; diff --git a/fs/xfs/linux-2.6/xfs_fs_subr.c b/fs/xfs/linux-2.6/xfs_fs_subr.c index 4fa4b1a5187..575f2a790f3 100644 --- a/fs/xfs/linux-2.6/xfs_fs_subr.c +++ b/fs/xfs/linux-2.6/xfs_fs_subr.c @@ -57,7 +57,7 @@ fs_tosspages( int fiopt) { vnode_t *vp = BHV_TO_VNODE(bdp); - struct inode *ip = LINVFS_GET_IP(vp); + struct inode *ip = vn_to_inode(vp); if (VN_CACHED(vp)) truncate_inode_pages(ip->i_mapping, first); @@ -76,7 +76,7 @@ fs_flushinval_pages( int fiopt) { vnode_t *vp = BHV_TO_VNODE(bdp); - struct inode *ip = LINVFS_GET_IP(vp); + struct inode *ip = vn_to_inode(vp); if (VN_CACHED(vp)) { filemap_write_and_wait(ip->i_mapping); @@ -98,7 +98,7 @@ fs_flush_pages( int fiopt) { vnode_t *vp = BHV_TO_VNODE(bdp); - struct inode *ip = LINVFS_GET_IP(vp); + struct inode *ip = vn_to_inode(vp); if (VN_CACHED(vp)) { filemap_fdatawrite(ip->i_mapping); diff --git a/fs/xfs/linux-2.6/xfs_ioctl.c b/fs/xfs/linux-2.6/xfs_ioctl.c index e435ad17419..84478491609 100644 --- a/fs/xfs/linux-2.6/xfs_ioctl.c +++ b/fs/xfs/linux-2.6/xfs_ioctl.c @@ -138,7 +138,7 @@ xfs_find_handle( } /* we need the vnode */ - vp = LINVFS_GET_VP(inode); + vp = vn_from_inode(inode); /* now we can grab the fsid */ memcpy(&handle.ha_fsid, vp->v_vfsp->vfs_altfsid, sizeof(xfs_fsid_t)); @@ -256,7 +256,7 @@ xfs_vget_fsop_handlereq( } vpp = XFS_ITOV(ip); - inodep = LINVFS_GET_IP(vpp); + inodep = vn_to_inode(vpp); xfs_iunlock(ip, XFS_ILOCK_SHARED); *vp = vpp; @@ -715,7 +715,7 @@ xfs_ioctl( xfs_inode_t *ip; xfs_mount_t *mp; - vp = LINVFS_GET_VP(inode); + vp = vn_from_inode(inode); vn_trace_entry(vp, "xfs_ioctl", (inst_t *)__return_address); @@ -1270,7 +1270,7 @@ xfs_ioc_xattr( } case XFS_IOC_GETVERSION: { - flags = LINVFS_GET_IP(vp)->i_generation; + flags = vn_to_inode(vp)->i_generation; if (copy_to_user(arg, &flags, sizeof(flags))) error = -EFAULT; break; diff --git a/fs/xfs/linux-2.6/xfs_ioctl32.c b/fs/xfs/linux-2.6/xfs_ioctl32.c index e9da0bde36a..b6321abd9a8 100644 --- a/fs/xfs/linux-2.6/xfs_ioctl32.c +++ b/fs/xfs/linux-2.6/xfs_ioctl32.c @@ -111,7 +111,7 @@ xfs_compat_ioctl(int mode, struct file *f, unsigned cmd, unsigned long arg) { int error; struct inode *inode = f->f_dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_to_inode(inode); switch (cmd) { case XFS_IOC_DIOINFO: diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c index 0a508580e57..64c3395074d 100644 --- a/fs/xfs/linux-2.6/xfs_iops.c +++ b/fs/xfs/linux-2.6/xfs_iops.c @@ -106,7 +106,7 @@ xfs_ichgtime( xfs_inode_t *ip, int flags) { - struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); + struct inode *inode = vn_to_inode(XFS_ITOV(ip)); timespec_t tv; nanotime(&tv); @@ -202,7 +202,7 @@ xfs_validate_fields( struct inode *ip, struct vattr *vattr) { - vnode_t *vp = LINVFS_GET_VP(ip); + vnode_t *vp = vn_from_inode(ip); int error; vattr->va_mask = XFS_AT_NLINK|XFS_AT_SIZE|XFS_AT_NBLOCKS; @@ -228,7 +228,7 @@ xfs_init_security( struct vnode *vp, struct inode *dir) { - struct inode *ip = LINVFS_GET_IP(vp); + struct inode *ip = vn_to_inode(vp); size_t length; void *value; char *name; @@ -277,7 +277,7 @@ xfs_cleanup_inode( * xfs_init_security we must back out. * ENOSPC can hit here, among other things. */ - teardown.d_inode = LINVFS_GET_IP(vp); + teardown.d_inode = vn_to_inode(vp); teardown.d_name = dentry->d_name; if (S_ISDIR(mode)) @@ -296,7 +296,7 @@ xfs_vn_mknod( { struct inode *ip; vattr_t vattr = { 0 }; - vnode_t *vp = NULL, *dvp = LINVFS_GET_VP(dir); + vnode_t *vp = NULL, *dvp = vn_from_inode(dir); xfs_acl_t *default_acl = NULL; attrexists_t test_default_acl = _ACL_DEFAULT_EXISTS; int error; @@ -359,7 +359,7 @@ xfs_vn_mknod( if (likely(!error)) { ASSERT(vp); - ip = LINVFS_GET_IP(vp); + ip = vn_to_inode(vp); if (S_ISCHR(mode) || S_ISBLK(mode)) ip->i_rdev = rdev; @@ -396,7 +396,7 @@ xfs_vn_lookup( struct dentry *dentry, struct nameidata *nd) { - struct vnode *vp = LINVFS_GET_VP(dir), *cvp; + struct vnode *vp = vn_from_inode(dir), *cvp; int error; if (dentry->d_name.len >= MAXNAMELEN) @@ -410,7 +410,7 @@ xfs_vn_lookup( return NULL; } - return d_splice_alias(LINVFS_GET_IP(cvp), dentry); + return d_splice_alias(vn_to_inode(cvp), dentry); } STATIC int @@ -429,8 +429,8 @@ xfs_vn_link( if (S_ISDIR(ip->i_mode)) return -EPERM; - tdvp = LINVFS_GET_VP(dir); - vp = LINVFS_GET_VP(ip); + tdvp = vn_from_inode(dir); + vp = vn_from_inode(ip); VOP_LINK(tdvp, vp, dentry, NULL, error); if (likely(!error)) { @@ -453,7 +453,7 @@ xfs_vn_unlink( int error; inode = dentry->d_inode; - dvp = LINVFS_GET_VP(dir); + dvp = vn_from_inode(dir); VOP_REMOVE(dvp, dentry, NULL, error); if (likely(!error)) { @@ -475,7 +475,7 @@ xfs_vn_symlink( vnode_t *cvp; /* used to lookup symlink to put in dentry */ int error; - dvp = LINVFS_GET_VP(dir); + dvp = vn_from_inode(dir); cvp = NULL; vattr.va_mode = S_IFLNK | @@ -487,7 +487,7 @@ xfs_vn_symlink( if (likely(!error && cvp)) { error = xfs_init_security(cvp, dir); if (likely(!error)) { - ip = LINVFS_GET_IP(cvp); + ip = vn_to_inode(cvp); d_instantiate(dentry, ip); xfs_validate_fields(dir, &vattr); xfs_validate_fields(ip, &vattr); @@ -502,7 +502,7 @@ xfs_vn_rmdir( struct dentry *dentry) { struct inode *inode = dentry->d_inode; - vnode_t *dvp = LINVFS_GET_VP(dir); + vnode_t *dvp = vn_from_inode(dir); vattr_t vattr; int error; @@ -527,8 +527,8 @@ xfs_vn_rename( vattr_t vattr; int error; - fvp = LINVFS_GET_VP(odir); - tvp = LINVFS_GET_VP(ndir); + fvp = vn_from_inode(odir); + tvp = vn_from_inode(ndir); VOP_RENAME(fvp, odentry, tvp, ndentry, NULL, error); if (likely(!error)) { @@ -573,7 +573,7 @@ xfs_vn_follow_link( return NULL; } - vp = LINVFS_GET_VP(dentry->d_inode); + vp = vn_from_inode(dentry->d_inode); iov.iov_base = link; iov.iov_len = MAXPATHLEN; @@ -616,7 +616,7 @@ xfs_vn_permission( int mode, struct nameidata *nd) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error; mode <<= 6; /* convert from linux to vnode access bits */ @@ -634,7 +634,7 @@ xfs_vn_getattr( struct kstat *stat) { struct inode *inode = dentry->d_inode; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error = 0; if (unlikely(vp->v_flag & VMODIFIED)) @@ -651,7 +651,7 @@ xfs_vn_setattr( { struct inode *inode = dentry->d_inode; unsigned int ia_valid = attr->ia_valid; - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); vattr_t vattr = { 0 }; int flags = 0; int error; @@ -717,7 +717,7 @@ xfs_vn_setxattr( size_t size, int flags) { - vnode_t *vp = LINVFS_GET_VP(dentry->d_inode); + vnode_t *vp = vn_from_inode(dentry->d_inode); char *attr = (char *)name; attrnames_t *namesp; int xflags = 0; @@ -747,7 +747,7 @@ xfs_vn_getxattr( void *data, size_t size) { - vnode_t *vp = LINVFS_GET_VP(dentry->d_inode); + vnode_t *vp = vn_from_inode(dentry->d_inode); char *attr = (char *)name; attrnames_t *namesp; int xflags = 0; @@ -776,7 +776,7 @@ xfs_vn_listxattr( char *data, size_t size) { - vnode_t *vp = LINVFS_GET_VP(dentry->d_inode); + vnode_t *vp = vn_from_inode(dentry->d_inode); int error, xflags = ATTR_KERNAMELS; ssize_t result; @@ -795,7 +795,7 @@ xfs_vn_removexattr( struct dentry *dentry, const char *name) { - vnode_t *vp = LINVFS_GET_VP(dentry->d_inode); + vnode_t *vp = vn_from_inode(dentry->d_inode); char *attr = (char *)name; attrnames_t *namesp; int xflags = 0; diff --git a/fs/xfs/linux-2.6/xfs_linux.h b/fs/xfs/linux-2.6/xfs_linux.h index bd88ccb0cad..1fe09f2d651 100644 --- a/fs/xfs/linux-2.6/xfs_linux.h +++ b/fs/xfs/linux-2.6/xfs_linux.h @@ -234,7 +234,7 @@ BUFFER_FNS(PrivateStart, unwritten); #define xfs_sort(a,n,s,fn) sort(a,n,s,fn,NULL) #define xfs_stack_trace() dump_stack() #define xfs_itruncate_data(ip, off) \ - (-vmtruncate(LINVFS_GET_IP(XFS_ITOV(ip)), (off))) + (-vmtruncate(vn_to_inode(XFS_ITOV(ip)), (off))) #define xfs_statvfs_fsid(statp, mp) \ ({ u64 id = huge_encode_dev((mp)->m_ddev_targp->bt_dev); \ __kernel_fsid_t *fsid = &(statp)->f_fsid; \ diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c index e0ab45fbfeb..2ede4bb7ecd 100644 --- a/fs/xfs/linux-2.6/xfs_lrw.c +++ b/fs/xfs/linux-2.6/xfs_lrw.c @@ -432,7 +432,7 @@ xfs_zero_eof( xfs_fsize_t isize, /* current inode size */ xfs_fsize_t end_size) /* terminal inode size */ { - struct inode *ip = LINVFS_GET_IP(vp); + struct inode *ip = vn_to_inode(vp); xfs_fileoff_t start_zero_fsb; xfs_fileoff_t end_zero_fsb; xfs_fileoff_t zero_count_fsb; diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c index 352aa3d40c1..d9d28a965ba 100644 --- a/fs/xfs/linux-2.6/xfs_super.c +++ b/fs/xfs/linux-2.6/xfs_super.c @@ -153,7 +153,7 @@ xfs_revalidate_inode( vnode_t *vp, xfs_inode_t *ip) { - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); inode->i_mode = ip->i_d.di_mode; inode->i_nlink = ip->i_d.di_nlink; @@ -210,7 +210,7 @@ xfs_initialize_vnode( int unlock) { xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); if (!inode_bhv->bd_vobj) { vp->v_vfsp = bhvtovfs(bdp); @@ -228,7 +228,7 @@ xfs_initialize_vnode( if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); xfs_set_inodeops(inode); - + ip->i_flags &= ~XFS_INEW; barrier(); @@ -340,14 +340,14 @@ xfs_fs_alloc_inode( vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); if (unlikely(!vp)) return NULL; - return LINVFS_GET_IP(vp); + return vn_to_inode(vp); } STATIC void xfs_fs_destroy_inode( struct inode *inode) { - kmem_zone_free(xfs_vnode_zone, LINVFS_GET_VP(inode)); + kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode)); } STATIC void @@ -358,7 +358,7 @@ xfs_fs_inode_init_once( { if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == SLAB_CTOR_CONSTRUCTOR) - inode_init_once(LINVFS_GET_IP((vnode_t *)vnode)); + inode_init_once(vn_to_inode((vnode_t *)vnode)); } STATIC int @@ -409,7 +409,7 @@ xfs_fs_write_inode( struct inode *inode, int sync) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error = 0, flags = FLUSH_INODE; if (vp) { @@ -432,7 +432,7 @@ STATIC void xfs_fs_clear_inode( struct inode *inode) { - vnode_t *vp = LINVFS_GET_VP(inode); + vnode_t *vp = vn_from_inode(inode); int error, cache; vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); @@ -511,7 +511,7 @@ void xfs_flush_inode( xfs_inode_t *ip) { - struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); + struct inode *inode = vn_to_inode(XFS_ITOV(ip)); struct vfs *vfs = XFS_MTOVFS(ip->i_mount); igrab(inode); @@ -536,7 +536,7 @@ void xfs_flush_device( xfs_inode_t *ip) { - struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); + struct inode *inode = vn_to_inode(XFS_ITOV(ip)); struct vfs *vfs = XFS_MTOVFS(ip->i_mount); igrab(inode); @@ -630,7 +630,7 @@ STATIC void xfs_fs_put_super( struct super_block *sb) { - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); int error; xfs_fs_stop_syncd(vfsp); @@ -650,7 +650,7 @@ STATIC void xfs_fs_write_super( struct super_block *sb) { - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); int error; if (sb->s_flags & MS_RDONLY) { @@ -667,7 +667,7 @@ xfs_fs_sync_super( struct super_block *sb, int wait) { - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); int error; int flags = SYNC_FSDATA; @@ -706,7 +706,7 @@ xfs_fs_statfs( struct super_block *sb, struct kstatfs *statp) { - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); int error; VFS_STATVFS(vfsp, statp, NULL, error); @@ -719,7 +719,7 @@ xfs_fs_remount( int *flags, char *options) { - vfs_t *vfsp = LINVFS_GET_VFS(sb); + vfs_t *vfsp = vfs_from_sb(sb); struct xfs_mount_args *args = xfs_args_allocate(sb); int error; @@ -734,7 +734,7 @@ STATIC void xfs_fs_lockfs( struct super_block *sb) { - VFS_FREEZE(LINVFS_GET_VFS(sb)); + VFS_FREEZE(vfs_from_sb(sb)); } STATIC int @@ -742,7 +742,7 @@ xfs_fs_show_options( struct seq_file *m, struct vfsmount *mnt) { - struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb); + struct vfs *vfsp = vfs_from_sb(mnt->mnt_sb); int error; VFS_SHOWARGS(vfsp, m, error); @@ -754,7 +754,7 @@ xfs_fs_quotasync( struct super_block *sb, int type) { - struct vfs *vfsp = LINVFS_GET_VFS(sb); + struct vfs *vfsp = vfs_from_sb(sb); int error; VFS_QUOTACTL(vfsp, Q_XQUOTASYNC, 0, (caddr_t)NULL, error); @@ -766,7 +766,7 @@ xfs_fs_getxstate( struct super_block *sb, struct fs_quota_stat *fqs) { - struct vfs *vfsp = LINVFS_GET_VFS(sb); + struct vfs *vfsp = vfs_from_sb(sb); int error; VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); @@ -779,7 +779,7 @@ xfs_fs_setxstate( unsigned int flags, int op) { - struct vfs *vfsp = LINVFS_GET_VFS(sb); + struct vfs *vfsp = vfs_from_sb(sb); int error; VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); @@ -793,7 +793,7 @@ xfs_fs_getxquota( qid_t id, struct fs_disk_quota *fdq) { - struct vfs *vfsp = LINVFS_GET_VFS(sb); + struct vfs *vfsp = vfs_from_sb(sb); int error, getmode; getmode = (type == USRQUOTA) ? Q_XGETQUOTA : @@ -809,7 +809,7 @@ xfs_fs_setxquota( qid_t id, struct fs_disk_quota *fdq) { - struct vfs *vfsp = LINVFS_GET_VFS(sb); + struct vfs *vfsp = vfs_from_sb(sb); int error, setmode; setmode = (type == USRQUOTA) ? Q_XSETQLIM : @@ -825,15 +825,11 @@ xfs_fs_fill_super( int silent) { vnode_t *rootvp; - struct vfs *vfsp = vfs_allocate(); + struct vfs *vfsp = vfs_allocate(sb); struct xfs_mount_args *args = xfs_args_allocate(sb); struct kstatfs statvfs; int error, error2; - vfsp->vfs_super = sb; - LINVFS_SET_VFS(sb, vfsp); - if (sb->s_flags & MS_RDONLY) - vfsp->vfs_flag |= VFS_RDONLY; bhv_insert_all_vfsops(vfsp); VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); @@ -871,7 +867,7 @@ xfs_fs_fill_super( if (error) goto fail_unmount; - sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp)); + sb->s_root = d_alloc_root(vn_to_inode(rootvp)); if (!sb->s_root) { error = ENOMEM; goto fail_vnrele; diff --git a/fs/xfs/linux-2.6/xfs_super.h b/fs/xfs/linux-2.6/xfs_super.h index 5e1b525cb1e..376b96cb513 100644 --- a/fs/xfs/linux-2.6/xfs_super.h +++ b/fs/xfs/linux-2.6/xfs_super.h @@ -98,11 +98,6 @@ extern void xfs_qm_exit(void); XFS_DMAPI_STRING \ XFS_DBG_STRING /* DBG must be last */ -#define LINVFS_GET_VFS(s) \ - (vfs_t *)((s)->s_fs_info) -#define LINVFS_SET_VFS(s, vfsp) \ - ((s)->s_fs_info = vfsp) - struct xfs_inode; struct xfs_mount; struct xfs_buftarg; diff --git a/fs/xfs/linux-2.6/xfs_vfs.c b/fs/xfs/linux-2.6/xfs_vfs.c index a6b4084bda8..6f7c9f7a862 100644 --- a/fs/xfs/linux-2.6/xfs_vfs.c +++ b/fs/xfs/linux-2.6/xfs_vfs.c @@ -227,7 +227,8 @@ vfs_freeze( } vfs_t * -vfs_allocate( void ) +vfs_allocate( + struct super_block *sb) { struct vfs *vfsp; @@ -236,9 +237,23 @@ vfs_allocate( void ) INIT_LIST_HEAD(&vfsp->vfs_sync_list); spin_lock_init(&vfsp->vfs_sync_lock); init_waitqueue_head(&vfsp->vfs_wait_single_sync_task); + + vfsp->vfs_super = sb; + sb->s_fs_info = vfsp; + + if (sb->s_flags & MS_RDONLY) + vfsp->vfs_flag |= VFS_RDONLY; + return vfsp; } +vfs_t * +vfs_from_sb( + struct super_block *sb) +{ + return (vfs_t *)sb->s_fs_info; +} + void vfs_deallocate( struct vfs *vfsp) diff --git a/fs/xfs/linux-2.6/xfs_vfs.h b/fs/xfs/linux-2.6/xfs_vfs.h index 57caf9eddee..8fed356db05 100644 --- a/fs/xfs/linux-2.6/xfs_vfs.h +++ b/fs/xfs/linux-2.6/xfs_vfs.h @@ -193,7 +193,8 @@ typedef struct bhv_vfsops { #define vfs_bhv_set_custom(b,o) ( (b)->bhv_custom = (void *)(o)) #define vfs_bhv_clr_custom(b) ( (b)->bhv_custom = NULL ) -extern vfs_t *vfs_allocate(void); +extern vfs_t *vfs_allocate(struct super_block *); +extern vfs_t *vfs_from_sb(struct super_block *); extern void vfs_deallocate(vfs_t *); extern void vfs_insertops(vfs_t *, bhv_vfsops_t *); extern void vfs_insertbhv(vfs_t *, bhv_desc_t *, vfsops_t *, void *); diff --git a/fs/xfs/linux-2.6/xfs_vnode.c b/fs/xfs/linux-2.6/xfs_vnode.c index 225e7dd8b21..d27c25b27cc 100644 --- a/fs/xfs/linux-2.6/xfs_vnode.c +++ b/fs/xfs/linux-2.6/xfs_vnode.c @@ -58,7 +58,7 @@ struct vnode * vn_initialize( struct inode *inode) { - struct vnode *vp = LINVFS_GET_VP(inode); + struct vnode *vp = vn_from_inode(inode); XFS_STATS_INC(vn_active); XFS_STATS_INC(vn_alloc); @@ -97,7 +97,7 @@ vn_revalidate_core( struct vnode *vp, vattr_t *vap) { - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); inode->i_mode = vap->va_mode; inode->i_nlink = vap->va_nlink; @@ -166,7 +166,7 @@ vn_hold( XFS_STATS_INC(vn_hold); VN_LOCK(vp); - inode = igrab(LINVFS_GET_IP(vp)); + inode = igrab(vn_to_inode(vp)); ASSERT(inode); VN_UNLOCK(vp, 0); diff --git a/fs/xfs/linux-2.6/xfs_vnode.h b/fs/xfs/linux-2.6/xfs_vnode.h index 0cf92ca80ce..06f5845e956 100644 --- a/fs/xfs/linux-2.6/xfs_vnode.h +++ b/fs/xfs/linux-2.6/xfs_vnode.h @@ -116,8 +116,14 @@ typedef enum { /* * Vnode to Linux inode mapping. */ -#define LINVFS_GET_VP(inode) ((vnode_t *)list_entry(inode, vnode_t, v_inode)) -#define LINVFS_GET_IP(vp) (&(vp)->v_inode) +static inline struct vnode *vn_from_inode(struct inode *inode) +{ + return (vnode_t *)list_entry(inode, vnode_t, v_inode); +} +static inline struct inode *vn_to_inode(struct vnode *vnode) +{ + return &vnode->v_inode; +} /* * Vnode flags. @@ -498,7 +504,7 @@ extern void vn_iowake(struct vnode *vp); static inline int vn_count(struct vnode *vp) { - return atomic_read(&LINVFS_GET_IP(vp)->i_count); + return atomic_read(&vn_to_inode(vp)->i_count); } /* @@ -512,16 +518,16 @@ extern vnode_t *vn_hold(struct vnode *); vn_trace_hold(vp, __FILE__, __LINE__, (inst_t *)__return_address)) #define VN_RELE(vp) \ (vn_trace_rele(vp, __FILE__, __LINE__, (inst_t *)__return_address), \ - iput(LINVFS_GET_IP(vp))) + iput(vn_to_inode(vp))) #else #define VN_HOLD(vp) ((void)vn_hold(vp)) -#define VN_RELE(vp) (iput(LINVFS_GET_IP(vp))) +#define VN_RELE(vp) (iput(vn_to_inode(vp))) #endif static inline struct vnode *vn_grab(struct vnode *vp) { - struct inode *inode = igrab(LINVFS_GET_IP(vp)); - return inode ? LINVFS_GET_VP(inode) : NULL; + struct inode *inode = igrab(vn_to_inode(vp)); + return inode ? vn_from_inode(inode) : NULL; } /* @@ -529,7 +535,7 @@ static inline struct vnode *vn_grab(struct vnode *vp) */ #define VNAME(dentry) ((char *) (dentry)->d_name.name) #define VNAMELEN(dentry) ((dentry)->d_name.len) -#define VNAME_TO_VNODE(dentry) (LINVFS_GET_VP((dentry)->d_inode)) +#define VNAME_TO_VNODE(dentry) (vn_from_inode((dentry)->d_inode)) /* * Vnode spinlock manipulation. @@ -558,12 +564,12 @@ static __inline__ void vn_flagclr(struct vnode *vp, uint flag) */ static inline void vn_mark_bad(struct vnode *vp) { - make_bad_inode(LINVFS_GET_IP(vp)); + make_bad_inode(vn_to_inode(vp)); } static inline int VN_BAD(struct vnode *vp) { - return is_bad_inode(LINVFS_GET_IP(vp)); + return is_bad_inode(vn_to_inode(vp)); } /* @@ -588,9 +594,9 @@ static inline void vn_atime_to_time_t(struct vnode *vp, time_t *tt) /* * Some useful predicates. */ -#define VN_MAPPED(vp) mapping_mapped(LINVFS_GET_IP(vp)->i_mapping) -#define VN_CACHED(vp) (LINVFS_GET_IP(vp)->i_mapping->nrpages) -#define VN_DIRTY(vp) mapping_tagged(LINVFS_GET_IP(vp)->i_mapping, \ +#define VN_MAPPED(vp) mapping_mapped(vn_to_inode(vp)->i_mapping) +#define VN_CACHED(vp) (vn_to_inode(vp)->i_mapping->nrpages) +#define VN_DIRTY(vp) mapping_tagged(vn_to_inode(vp)->i_mapping, \ PAGECACHE_TAG_DIRTY) #define VMODIFY(vp) VN_FLAGSET(vp, VMODIFIED) #define VUNMODIFY(vp) VN_FLAGCLR(vp, VMODIFIED) diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c index e5e91e9c7e8..acf1b7c9f1d 100644 --- a/fs/xfs/xfs_attr.c +++ b/fs/xfs/xfs_attr.c @@ -2522,7 +2522,7 @@ attr_user_capable( struct vnode *vp, cred_t *cred) { - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) return -EPERM; @@ -2540,7 +2540,7 @@ attr_trusted_capable( struct vnode *vp, cred_t *cred) { - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) return -EPERM; diff --git a/fs/xfs/xfs_dfrag.c b/fs/xfs/xfs_dfrag.c index c6191d00ad2..4968a6358e6 100644 --- a/fs/xfs/xfs_dfrag.c +++ b/fs/xfs/xfs_dfrag.c @@ -83,7 +83,7 @@ xfs_swapext( /* Pull information for the target fd */ if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) || - ((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) { + ((vp = vn_from_inode(fp->f_dentry->d_inode)) == NULL)) { error = XFS_ERROR(EINVAL); goto error0; } @@ -95,7 +95,7 @@ xfs_swapext( } if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) || - ((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) { + ((tvp = vn_from_inode(tfp->f_dentry->d_inode)) == NULL)) { error = XFS_ERROR(EINVAL); goto error0; } diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c index 8e380a1fb79..3ce35a6f700 100644 --- a/fs/xfs/xfs_iget.c +++ b/fs/xfs/xfs_iget.c @@ -258,7 +258,7 @@ again: goto finish_inode; } else if (vp != inode_vp) { - struct inode *inode = LINVFS_GET_IP(inode_vp); + struct inode *inode = vn_to_inode(inode_vp); /* The inode is being torn down, pause and * try again. @@ -495,7 +495,7 @@ retry: if ((inode = iget_locked(XFS_MTOVFS(mp)->vfs_super, ino))) { xfs_inode_t *ip; - vp = LINVFS_GET_VP(inode); + vp = vn_from_inode(inode); if (inode->i_state & I_NEW) { vn_initialize(inode); error = xfs_iget_core(vp, mp, tp, ino, flags, @@ -617,7 +617,7 @@ xfs_iput_new(xfs_inode_t *ip, uint lock_flags) { vnode_t *vp = XFS_ITOV(ip); - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); vn_trace_entry(vp, "xfs_iput_new", (inst_t *)__return_address); diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index a16df2d435f..86a8451d10f 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2723,7 +2723,7 @@ xfs_iunpin( /* make sync come back and flush this inode */ if (vp) { - struct inode *inode = LINVFS_GET_IP(vp); + struct inode *inode = vn_to_inode(vp); if (!(inode->i_state & I_NEW)) mark_inode_dirty_sync(inode); @@ -3519,7 +3519,7 @@ xfs_iaccess( { int error; mode_t orgmode = mode; - struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); + struct inode *inode = vn_to_inode(XFS_ITOV(ip)); if (mode & S_IWUSR) { umode_t imode = inode->i_mode; diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c index eaab355f5a8..8b5a44fe286 100644 --- a/fs/xfs/xfs_vnodeops.c +++ b/fs/xfs/xfs_vnodeops.c @@ -1556,7 +1556,7 @@ xfs_release( if ((error = xfs_inactive_free_eofblocks(mp, ip))) return error; /* Update linux inode block count after free above */ - LINVFS_GET_IP(vp)->i_blocks = XFS_FSB_TO_BB(mp, + vn_to_inode(vp)->i_blocks = XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); } } @@ -1637,7 +1637,7 @@ xfs_inactive( if ((error = xfs_inactive_free_eofblocks(mp, ip))) return VN_INACTIVE_CACHE; /* Update linux inode block count after free above */ - LINVFS_GET_IP(vp)->i_blocks = XFS_FSB_TO_BB(mp, + vn_to_inode(vp)->i_blocks = XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); } goto out; -- cgit v1.2.3-70-g09d2 From b12dd34298cf0cff9f337f667045335140873039 Mon Sep 17 00:00:00 2001 From: Nathan Scott Date: Fri, 17 Mar 2006 17:26:04 +1100 Subject: [XFS] Fix an infinite loop issue in bulkstat when a corrupt inode is detected. Thanks to Roger Willcocks. SGI-PV: 951054 SGI-Modid: xfs-linux-melb:xfs-kern:25477a Signed-off-by: Nathan Scott --- fs/xfs/xfs_inode.c | 28 ++++++++++++++-------------- fs/xfs/xfs_inode.h | 7 ++++--- fs/xfs/xfs_itable.c | 5 ++++- fs/xfs/xfs_log_recover.c | 2 +- fs/xfs/xfs_vfsops.c | 2 +- 5 files changed, 24 insertions(+), 20 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 86a8451d10f..b1e95707e7c 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -253,7 +253,8 @@ xfs_itobp( xfs_inode_t *ip, xfs_dinode_t **dipp, xfs_buf_t **bpp, - xfs_daddr_t bno) + xfs_daddr_t bno, + uint imap_flags) { xfs_buf_t *bp; int error; @@ -269,10 +270,9 @@ xfs_itobp( * inode on disk. */ imap.im_blkno = bno; - error = xfs_imap(mp, tp, ip->i_ino, &imap, XFS_IMAP_LOOKUP); - if (error != 0) { + if ((error = xfs_imap(mp, tp, ip->i_ino, &imap, + XFS_IMAP_LOOKUP | imap_flags))) return error; - } /* * If the inode number maps to a block outside the bounds @@ -336,9 +336,10 @@ xfs_itobp( * (if DEBUG kernel) or the first inode in the buffer, otherwise. */ #ifdef DEBUG - ni = BBTOB(imap.im_len) >> mp->m_sb.sb_inodelog; + ni = (imap_flags & XFS_IMAP_BULKSTAT) ? 0 : + (BBTOB(imap.im_len) >> mp->m_sb.sb_inodelog); #else - ni = 1; + ni = (imap_flags & XFS_IMAP_BULKSTAT) ? 0 : 1; #endif for (i = 0; i < ni; i++) { int di_ok; @@ -868,9 +869,8 @@ xfs_iread( * return NULL as well. Set i_blkno to 0 so that xfs_itobp() will * know that this is a new incore inode. */ - error = xfs_itobp(mp, tp, ip, &dip, &bp, bno); - - if (error != 0) { + error = xfs_itobp(mp, tp, ip, &dip, &bp, bno, 0); + if (error) { kmem_zone_free(xfs_inode_zone, ip); return error; } @@ -1895,7 +1895,7 @@ xfs_iunlink( * Here we put the head pointer into our next pointer, * and then we fall through to point the head at us. */ - error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0); + error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0, 0); if (error) { return error; } @@ -2004,7 +2004,7 @@ xfs_iunlink_remove( * of dealing with the buffer when there is no need to * change it. */ - error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0); + error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0, 0); if (error) { cmn_err(CE_WARN, "xfs_iunlink_remove: xfs_itobp() returned an error %d on %s. Returning error.", @@ -2066,7 +2066,7 @@ xfs_iunlink_remove( * Now last_ibp points to the buffer previous to us on * the unlinked list. Pull us from the list. */ - error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0); + error = xfs_itobp(mp, tp, ip, &dip, &ibp, 0, 0); if (error) { cmn_err(CE_WARN, "xfs_iunlink_remove: xfs_itobp() returned an error %d on %s. Returning error.", @@ -3023,8 +3023,8 @@ xfs_iflush( /* * Get the buffer containing the on-disk inode. */ - error = xfs_itobp(mp, NULL, ip, &dip, &bp, 0); - if (error != 0) { + error = xfs_itobp(mp, NULL, ip, &dip, &bp, 0, 0); + if (error) { xfs_ifunlock(ip); return error; } diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index 006396764cb..39ef9c36ea5 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -95,9 +95,10 @@ typedef struct xfs_ifork { #define XFS_IFEXTIREC 0x08 /* Indirection array of extent blocks */ /* - * Flags for xfs_imap() and xfs_dilocate(). + * Flags for xfs_itobp(), xfs_imap() and xfs_dilocate(). */ -#define XFS_IMAP_LOOKUP 0x1 +#define XFS_IMAP_LOOKUP 0x1 +#define XFS_IMAP_BULKSTAT 0x2 #ifdef __KERNEL__ struct bhv_desc; @@ -421,7 +422,7 @@ int xfs_finish_reclaim_all(struct xfs_mount *, int); */ int xfs_itobp(struct xfs_mount *, struct xfs_trans *, xfs_inode_t *, xfs_dinode_t **, struct xfs_buf **, - xfs_daddr_t); + xfs_daddr_t, uint); int xfs_iread(struct xfs_mount *, struct xfs_trans *, xfs_ino_t, xfs_inode_t **, xfs_daddr_t); int xfs_iread_extents(struct xfs_trans *, xfs_inode_t *, int); diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c index c59450e1be4..32247b6bfee 100644 --- a/fs/xfs/xfs_itable.c +++ b/fs/xfs/xfs_itable.c @@ -562,7 +562,8 @@ xfs_bulkstat( if (bp) xfs_buf_relse(bp); error = xfs_itobp(mp, NULL, ip, - &dip, &bp, bno); + &dip, &bp, bno, + XFS_IMAP_BULKSTAT); if (!error) clustidx = ip->i_boffset / mp->m_sb.sb_inodesize; kmem_zone_free(xfs_inode_zone, ip); @@ -570,6 +571,8 @@ xfs_bulkstat( mp, XFS_ERRTAG_BULKSTAT_READ_CHUNK, XFS_RANDOM_BULKSTAT_READ_CHUNK)) { bp = NULL; + ubleft = 0; + rval = error; break; } } diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c index 7d46cbd6a07..add13f507ed 100644 --- a/fs/xfs/xfs_log_recover.c +++ b/fs/xfs/xfs_log_recover.c @@ -3249,7 +3249,7 @@ xlog_recover_process_iunlinks( * next inode in the bucket. */ error = xfs_itobp(mp, NULL, ip, &dip, - &ibp, 0); + &ibp, 0, 0); ASSERT(error || (dip != NULL)); } diff --git a/fs/xfs/xfs_vfsops.c b/fs/xfs/xfs_vfsops.c index 5dd84fe609c..811a4261fa2 100644 --- a/fs/xfs/xfs_vfsops.c +++ b/fs/xfs/xfs_vfsops.c @@ -1229,7 +1229,7 @@ xfs_sync_inodes( xfs_iunlock(ip, XFS_ILOCK_SHARED); error = xfs_itobp(mp, NULL, ip, - &dip, &bp, 0); + &dip, &bp, 0, 0); if (!error) { xfs_buf_relse(bp); } else { -- cgit v1.2.3-70-g09d2 From 3b244aa81ecb06859e0c16abf4c26404c1d86591 Mon Sep 17 00:00:00 2001 From: Nathan Scott Date: Fri, 17 Mar 2006 17:29:25 +1100 Subject: [XFS] endianess annotations for xfs_attr_shortform_t SGI-PV: 943272 SGI-Modid: xfs-linux-melb:xfs-kern:25501a Signed-off-by: Christoph Hellwig Signed-off-by: Nathan Scott --- fs/xfs/xfs_attr_leaf.c | 55 +++++++++++++++++++++++--------------------------- fs/xfs/xfs_attr_sf.h | 8 ++++---- fs/xfs/xfs_inode.c | 2 +- 3 files changed, 30 insertions(+), 35 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c index 531417b2c74..37d2e10f874 100644 --- a/fs/xfs/xfs_attr_leaf.c +++ b/fs/xfs/xfs_attr_leaf.c @@ -194,7 +194,7 @@ xfs_attr_shortform_create(xfs_da_args_t *args) xfs_idata_realloc(dp, sizeof(*hdr), XFS_ATTR_FORK); hdr = (xfs_attr_sf_hdr_t *)ifp->if_u1.if_data; hdr->count = 0; - INT_SET(hdr->totsize, ARCH_CONVERT, sizeof(*hdr)); + hdr->totsize = cpu_to_be16(sizeof(*hdr)); xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_ADATA); } @@ -224,8 +224,7 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff) ASSERT(ifp->if_flags & XFS_IFINLINE); sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data; sfe = &sf->list[0]; - for (i = 0; i < INT_GET(sf->hdr.count, ARCH_CONVERT); - sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) { + for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) { #ifdef DEBUG if (sfe->namelen != args->namelen) continue; @@ -248,13 +247,13 @@ xfs_attr_shortform_add(xfs_da_args_t *args, int forkoff) sfe = (xfs_attr_sf_entry_t *)((char *)sf + offset); sfe->namelen = args->namelen; - INT_SET(sfe->valuelen, ARCH_CONVERT, args->valuelen); + sfe->valuelen = args->valuelen; sfe->flags = (args->flags & ATTR_SECURE) ? XFS_ATTR_SECURE : ((args->flags & ATTR_ROOT) ? XFS_ATTR_ROOT : 0); memcpy(sfe->nameval, args->name, args->namelen); memcpy(&sfe->nameval[args->namelen], args->value, args->valuelen); - INT_MOD(sf->hdr.count, ARCH_CONVERT, 1); - INT_MOD(sf->hdr.totsize, ARCH_CONVERT, size); + sf->hdr.count++; + be16_add(&sf->hdr.totsize, size); xfs_trans_log_inode(args->trans, dp, XFS_ILOG_CORE | XFS_ILOG_ADATA); xfs_sbversion_add_attr2(mp, args->trans); @@ -277,7 +276,7 @@ xfs_attr_shortform_remove(xfs_da_args_t *args) base = sizeof(xfs_attr_sf_hdr_t); sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data; sfe = &sf->list[0]; - end = INT_GET(sf->hdr.count, ARCH_CONVERT); + end = sf->hdr.count; for (i = 0; i < end; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), base += size, i++) { size = XFS_ATTR_SF_ENTSIZE(sfe); @@ -300,11 +299,11 @@ xfs_attr_shortform_remove(xfs_da_args_t *args) * Fix up the attribute fork data, covering the hole */ end = base + size; - totsize = INT_GET(sf->hdr.totsize, ARCH_CONVERT); + totsize = be16_to_cpu(sf->hdr.totsize); if (end != totsize) memmove(&((char *)sf)[base], &((char *)sf)[end], totsize - end); - INT_MOD(sf->hdr.count, ARCH_CONVERT, -1); - INT_MOD(sf->hdr.totsize, ARCH_CONVERT, -size); + sf->hdr.count--; + be16_add(&sf->hdr.totsize, -size); /* * Fix up the start offset of the attribute fork @@ -360,7 +359,7 @@ xfs_attr_shortform_lookup(xfs_da_args_t *args) ASSERT(ifp->if_flags & XFS_IFINLINE); sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data; sfe = &sf->list[0]; - for (i = 0; i < INT_GET(sf->hdr.count, ARCH_CONVERT); + for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) { if (sfe->namelen != args->namelen) continue; @@ -391,7 +390,7 @@ xfs_attr_shortform_getvalue(xfs_da_args_t *args) ASSERT(args->dp->i_d.di_aformat == XFS_IFINLINE); sf = (xfs_attr_shortform_t *)args->dp->i_afp->if_u1.if_data; sfe = &sf->list[0]; - for (i = 0; i < INT_GET(sf->hdr.count, ARCH_CONVERT); + for (i = 0; i < sf->hdr.count; sfe = XFS_ATTR_SF_NEXTENTRY(sfe), i++) { if (sfe->namelen != args->namelen) continue; @@ -404,14 +403,14 @@ xfs_attr_shortform_getvalue(xfs_da_args_t *args) ((sfe->flags & XFS_ATTR_ROOT) != 0)) continue; if (args->flags & ATTR_KERNOVAL) { - args->valuelen = INT_GET(sfe->valuelen, ARCH_CONVERT); + args->valuelen = sfe->valuelen; return(XFS_ERROR(EEXIST)); } - if (args->valuelen < INT_GET(sfe->valuelen, ARCH_CONVERT)) { - args->valuelen = INT_GET(sfe->valuelen, ARCH_CONVERT); + if (args->valuelen < sfe->valuelen) { + args->valuelen = sfe->valuelen; return(XFS_ERROR(ERANGE)); } - args->valuelen = INT_GET(sfe->valuelen, ARCH_CONVERT); + args->valuelen = sfe->valuelen; memcpy(args->value, &sfe->nameval[args->namelen], args->valuelen); return(XFS_ERROR(EEXIST)); @@ -438,7 +437,7 @@ xfs_attr_shortform_to_leaf(xfs_da_args_t *args) dp = args->dp; ifp = dp->i_afp; sf = (xfs_attr_shortform_t *)ifp->if_u1.if_data; - size = INT_GET(sf->hdr.totsize, ARCH_CONVERT); + size = be16_to_cpu(sf->hdr.totsize); tmpbuffer = kmem_alloc(size, KM_SLEEP); ASSERT(tmpbuffer != NULL); memcpy(tmpbuffer, ifp->if_u1.if_data, size); @@ -481,11 +480,11 @@ xfs_attr_shortform_to_leaf(xfs_da_args_t *args) nargs.oknoent = 1; sfe = &sf->list[0]; - for (i = 0; i < INT_GET(sf->hdr.count, ARCH_CONVERT); i++) { + for (i = 0; i < sf->hdr.count; i++) { nargs.name = (char *)sfe->nameval; nargs.namelen = sfe->namelen; nargs.value = (char *)&sfe->nameval[nargs.namelen]; - nargs.valuelen = INT_GET(sfe->valuelen, ARCH_CONVERT); + nargs.valuelen = sfe->valuelen; nargs.hashval = xfs_da_hashname((char *)sfe->nameval, sfe->namelen); nargs.flags = (sfe->flags & XFS_ATTR_SECURE) ? ATTR_SECURE : @@ -560,10 +559,8 @@ xfs_attr_shortform_list(xfs_attr_list_context_t *context) * If the buffer is large enough, do not bother with sorting. * Note the generous fudge factor of 16 overhead bytes per entry. */ - if ((dp->i_afp->if_bytes + INT_GET(sf->hdr.count, ARCH_CONVERT) * 16) - < context->bufsize) { - for (i = 0, sfe = &sf->list[0]; - i < INT_GET(sf->hdr.count, ARCH_CONVERT); i++) { + if ((dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize) { + for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { attrnames_t *namesp; if (((context->flags & ATTR_SECURE) != 0) != @@ -584,14 +581,13 @@ xfs_attr_shortform_list(xfs_attr_list_context_t *context) if (context->flags & ATTR_KERNOVAL) { ASSERT(context->flags & ATTR_KERNAMELS); context->count += namesp->attr_namelen + - INT_GET(sfe->namelen, ARCH_CONVERT) + 1; + sfe->namelen + 1; } else { if (xfs_attr_put_listent(context, namesp, (char *)sfe->nameval, (int)sfe->namelen, - (int)INT_GET(sfe->valuelen, - ARCH_CONVERT))) + (int)sfe->valuelen)) break; } sfe = XFS_ATTR_SF_NEXTENTRY(sfe); @@ -603,7 +599,7 @@ xfs_attr_shortform_list(xfs_attr_list_context_t *context) /* * It didn't all fit, so we have to sort everything on hashval. */ - sbsize = INT_GET(sf->hdr.count, ARCH_CONVERT) * sizeof(*sbuf); + sbsize = sf->hdr.count * sizeof(*sbuf); sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP); /* @@ -611,8 +607,7 @@ xfs_attr_shortform_list(xfs_attr_list_context_t *context) * the relevant info from only those that match into a buffer. */ nsbuf = 0; - for (i = 0, sfe = &sf->list[0]; - i < INT_GET(sf->hdr.count, ARCH_CONVERT); i++) { + for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) { if (unlikely( ((char *)sfe < (char *)sf) || ((char *)sfe >= ((char *)sf + dp->i_afp->if_bytes)))) { @@ -696,7 +691,7 @@ xfs_attr_shortform_list(xfs_attr_list_context_t *context) } else { if (xfs_attr_put_listent(context, namesp, sbp->name, sbp->namelen, - INT_GET(sbp->valuelen, ARCH_CONVERT))) + sbp->valuelen)) break; } cursor->offset++; diff --git a/fs/xfs/xfs_attr_sf.h b/fs/xfs/xfs_attr_sf.h index ffed6ca81a5..f67f917803b 100644 --- a/fs/xfs/xfs_attr_sf.h +++ b/fs/xfs/xfs_attr_sf.h @@ -32,8 +32,8 @@ struct xfs_inode; */ typedef struct xfs_attr_shortform { struct xfs_attr_sf_hdr { /* constant-structure header block */ - __uint16_t totsize; /* total bytes in shortform list */ - __uint8_t count; /* count of active entries */ + __be16 totsize; /* total bytes in shortform list */ + __u8 count; /* count of active entries */ } hdr; struct xfs_attr_sf_entry { __uint8_t namelen; /* actual length of name (no NULL) */ @@ -66,8 +66,8 @@ typedef struct xfs_attr_sf_sort { #define XFS_ATTR_SF_NEXTENTRY(sfep) /* next entry in struct */ \ ((xfs_attr_sf_entry_t *)((char *)(sfep) + XFS_ATTR_SF_ENTSIZE(sfep))) #define XFS_ATTR_SF_TOTSIZE(dp) /* total space in use */ \ - (INT_GET(((xfs_attr_shortform_t *) \ - ((dp)->i_afp->if_u1.if_data))->hdr.totsize, ARCH_CONVERT)) + (be16_to_cpu(((xfs_attr_shortform_t *) \ + ((dp)->i_afp->if_u1.if_data))->hdr.totsize)) #if defined(XFS_ATTR_TRACE) /* diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index b1e95707e7c..7d0ded05a46 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -506,7 +506,7 @@ xfs_iformat( switch (INT_GET(dip->di_core.di_aformat, ARCH_CONVERT)) { case XFS_DINODE_FMT_LOCAL: atp = (xfs_attr_shortform_t *)XFS_DFORK_APTR(dip); - size = (int)INT_GET(atp->hdr.totsize, ARCH_CONVERT); + size = be16_to_cpu(atp->hdr.totsize); error = xfs_iformat_local(ip, dip, XFS_ATTR_FORK, size); break; case XFS_DINODE_FMT_EXTENTS: -- cgit v1.2.3-70-g09d2 From f1fdc848aab7fb95b32e058b7f06cc07912b3734 Mon Sep 17 00:00:00 2001 From: Yingping Lu Date: Wed, 22 Mar 2006 12:44:15 +1100 Subject: [XFS] Fixing KDB's xrwtrc command, also added the current process id into the trace. SGI-PV: 948300 SGI-Modid: xfs-linux-melb:xfs-kern:208069a Signed-off-by: Yingping Lu Signed-off-by: Nathan Scott --- fs/xfs/linux-2.6/xfs_aops.c | 2 +- fs/xfs/linux-2.6/xfs_lrw.c | 4 ++-- fs/xfs/xfs_inode.c | 8 ++++---- fs/xfs/xfs_iomap.c | 2 +- 4 files changed, 8 insertions(+), 8 deletions(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c index d9bf130c63b..965757cf15e 100644 --- a/fs/xfs/linux-2.6/xfs_aops.c +++ b/fs/xfs/linux-2.6/xfs_aops.c @@ -103,7 +103,7 @@ xfs_page_trace( (void *)((unsigned long)delalloc), (void *)((unsigned long)unmapped), (void *)((unsigned long)unwritten), - (void *)NULL, + (void *)((unsigned long)current_pid()), (void *)NULL); } #else diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c index 2a936bd38ce..0169360475c 100644 --- a/fs/xfs/linux-2.6/xfs_lrw.c +++ b/fs/xfs/linux-2.6/xfs_lrw.c @@ -83,7 +83,7 @@ xfs_rw_enter_trace( (void *)((unsigned long)ioflags), (void *)((unsigned long)((io->io_new_size >> 32) & 0xffffffff)), (void *)((unsigned long)(io->io_new_size & 0xffffffff)), - (void *)NULL, + (void *)((unsigned long)current_pid()), (void *)NULL, (void *)NULL, (void *)NULL, @@ -113,7 +113,7 @@ xfs_inval_cached_trace( (void *)((unsigned long)(first & 0xffffffff)), (void *)((unsigned long)((last >> 32) & 0xffffffff)), (void *)((unsigned long)(last & 0xffffffff)), - (void *)NULL, + (void *)((unsigned long)current_pid()), (void *)NULL, (void *)NULL, (void *)NULL, diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 7d0ded05a46..2424a477794 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -1372,10 +1372,10 @@ xfs_itrunc_trace( (void*)(unsigned long)((toss_finish >> 32) & 0xffffffff), (void*)(unsigned long)(toss_finish & 0xffffffff), (void*)(unsigned long)current_cpu(), - (void*)0, - (void*)0, - (void*)0, - (void*)0); + (void*)(unsigned long)current_pid(), + (void*)NULL, + (void*)NULL, + (void*)NULL); } #else #define xfs_itrunc_trace(tag, ip, flag, new_size, toss_start, toss_finish) diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c index 788917f355c..d5dfedcb892 100644 --- a/fs/xfs/xfs_iomap.c +++ b/fs/xfs/xfs_iomap.c @@ -76,7 +76,7 @@ xfs_iomap_enter_trace( (void *)((unsigned long)count), (void *)((unsigned long)((io->io_new_size >> 32) & 0xffffffff)), (void *)((unsigned long)(io->io_new_size & 0xffffffff)), - (void *)NULL, + (void *)((unsigned long)current_pid()), (void *)NULL, (void *)NULL, (void *)NULL, -- cgit v1.2.3-70-g09d2 From 9fa8046f50bcb88ab9183ee1f22de5adc42bf92a Mon Sep 17 00:00:00 2001 From: Yingping Lu Date: Wed, 22 Mar 2006 12:44:35 +1100 Subject: [XFS] Fixing the error caused by the conflict between DIO Write's conversion and concurrent truncate operations. Use vn_iowait to wait for the completion of any pending DIOs. Since the truncate requires exclusive IOLOCK, so this blocks any further DIO operations since DIO write also needs exclusive IOBLOCK. This serves as a barrier and prevent any potential starvation. SGI-PV: 947420 SGI-Modid: xfs-linux-melb:xfs-kern:208088a Signed-off-by: Yingping Lu Signed-off-by: Nathan Scott --- fs/xfs/xfs_inode.c | 3 +++ fs/xfs/xfs_vnodeops.c | 5 ++++- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 2424a477794..8d2b36879f0 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -1420,6 +1420,9 @@ xfs_itruncate_start( mp = ip->i_mount; vp = XFS_ITOV(ip); + + vn_iowait(vp); /* wait for the completion of any pending DIOs */ + /* * Call VOP_TOSS_PAGES() or VOP_FLUSHINVAL_PAGES() to get rid of pages and buffers * overlapping the region being removed. We have to use diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c index 8b5a44fe286..697bf22a84f 100644 --- a/fs/xfs/xfs_vnodeops.c +++ b/fs/xfs/xfs_vnodeops.c @@ -615,6 +615,7 @@ xfs_setattr( code = xfs_igrow_start(ip, vap->va_size, credp); } xfs_iunlock(ip, XFS_ILOCK_EXCL); + vn_iowait(vp); /* wait for the completion of any pending DIOs */ if (!code) code = xfs_itruncate_data(ip, vap->va_size); if (code) { @@ -4310,8 +4311,10 @@ xfs_free_file_space( ASSERT(attr_flags & ATTR_NOLOCK ? attr_flags & ATTR_DMI : 1); if (attr_flags & ATTR_NOLOCK) need_iolock = 0; - if (need_iolock) + if (need_iolock) { xfs_ilock(ip, XFS_IOLOCK_EXCL); + vn_iowait(vp); /* wait for the completion of any pending DIOs */ + } rounding = MAX((__uint8_t)(1 << mp->m_sb.sb_blocklog), (__uint8_t)NBPP); -- cgit v1.2.3-70-g09d2 From 38e2299a641d93d029eb559e096648ab75a22be2 Mon Sep 17 00:00:00 2001 From: David Chinner Date: Wed, 22 Mar 2006 12:47:15 +1100 Subject: [XFS] Explain the race closed by the addition of vn_iowait() to the start of xfs_itruncate_start(). SGI-PV: 947420 SGI-Modid: xfs-linux-melb:xfs-kern:25527a Signed-off-by: David Chinner Signed-off-by: Nathan Scott --- fs/xfs/xfs_inode.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'fs/xfs/xfs_inode.c') diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 8d2b36879f0..88a517fad07 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -1393,6 +1393,16 @@ xfs_itrunc_trace( * calling into the buffer/page cache code and we can't hold the * inode lock when we do so. * + * We need to wait for any direct I/Os in flight to complete before we + * proceed with the truncate. This is needed to prevent the extents + * being read or written by the direct I/Os from being removed while the + * I/O is in flight as there is no other method of synchronising + * direct I/O with the truncate operation. Also, because we hold + * the IOLOCK in exclusive mode, we prevent new direct I/Os from being + * started until the truncate completes and drops the lock. Essentially, + * the vn_iowait() call forms an I/O barrier that provides strict ordering + * between direct I/Os and the truncate operation. + * * The flags parameter can have either the value XFS_ITRUNC_DEFINITE * or XFS_ITRUNC_MAYBE. The XFS_ITRUNC_MAYBE value should be used * in the case that the caller is locking things out of order and -- cgit v1.2.3-70-g09d2