Re: [EXT4 set 2][PATCH 2/5] cleanups: Add extent sanity checks
On Tue, 2007-07-10 at 16:30 -0700, Andrew Morton wrote: > On Sun, 01 Jul 2007 03:36:22 -0400 > Mingming Cao <[EMAIL PROTECTED]> wrote: > > > with the patch all headers are checked. the code should become > > more resistant to on-disk corruptions. needless BUG_ON() have > > been removed. please, review for inclusion. > > > > ... > > > @@ -269,6 +239,70 @@ > > return size; > > } > > > > +static inline int > > +ext4_ext_max_entries(struct inode *inode, int depth) > > Please remove the `inline'. > done > `inline' is usually wrong. It is wrong here. If this function has a > single callsite (it has) then the compiler will inline it. If the function > later gains more callsites then the compiler will know (correctly) not to > inline it. > > We hope. If we find the compiler still inlines a function as large as this > one then we'd need to use noinline and complain at the gcc developers. > > > +{ > > + int max; > > + > > + if (depth == ext_depth(inode)) { > > + if (depth == 0) > > + max = ext4_ext_space_root(inode); > > + else > > + max = ext4_ext_space_root_idx(inode); > > + } else { > > + if (depth == 0) > > + max = ext4_ext_space_block(inode); > > + else > > + max = ext4_ext_space_block_idx(inode); > > + } > > + > > + return max; > > +} > > + > > +static int __ext4_ext_check_header(const char *function, struct inode > > *inode, > > + struct ext4_extent_header *eh, > > + int depth) > > +{ > > + const char *error_msg = NULL; > > Unneeded initialisation. > done > > + int max = 0; > > + > > + if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) { > > + error_msg = "invalid magic"; > > + goto corrupted; > > + } > > + if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) { > > + error_msg = "unexpected eh_depth"; > > + goto corrupted; > > + } > > + if (unlikely(eh->eh_max == 0)) { > > + error_msg = "invalid eh_max"; > > + goto corrupted; > > + } > > + max = ext4_ext_max_entries(inode, depth); > > + if (unlikely(le16_to_cpu(eh->eh_max) > max)) { > > + error_msg = "too large eh_max"; > > + goto corrupted; > > + } > > + if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) { > > + error_msg = "invalid eh_entries"; > > + goto corrupted; > > + } > > + return 0; > > + > > +corrupted: > > + ext4_error(inode->i_sb, function, > > + "bad header in inode #%lu: %s - magic %x, " > > + "entries %u, max %u(%u), depth %u(%u)", > > + inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic), > > + le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max), > > + max, le16_to_cpu(eh->eh_depth), depth); > > + > > + return -EIO; > > +} > > + > > > > ... > > > > + i = depth = ext_depth(inode); > > > > Mulitple assignments like this are somewhat unpopular from the coding-style > POV. > okay. > We have a local variable called `i' which is not used as a simple counter > and which has no explanatory comment. This is plain bad. Please find a > better name for this variable. Review the code for other such lack of > clarity - I'm sure there's more. > "i" is is loop counter. I have moved the i= depth to before the while loop. > > > - BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max)); > > - BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC); > > Yeah, this patch improves things a lot. > > How well-tested is it? Have any of these errors actually been triggered? > > > path[i].p_hdr = ext_block_hdr(path[i].p_bh); > > - if (ext4_ext_check_header(__FUNCTION__, inode, > > - path[i].p_hdr)) { > > - err = -EIO; > > - goto out; > > - } > > } > > > > - BUG_ON(le16_to_cpu(path[i].p_hdr->eh_entries) > > - > le16_to_cpu(path[i].p_hdr->eh_max)); > > - BUG_ON(path[i].p_hdr->eh_magic != EXT4_EXT_MAGIC); > > - > > if (!path[i].p_idx) { > > /* this level hasn't been touched yet */ > > path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); > > @@ -1873,17 +1890,24 @@ > > i, EXT_FIRST_INDEX(path[i].p_hdr), > > path[i].p_idx); > > if (ext4_ext_more_to_rm(path + i)) { > > + struct buffer_head *bh; > > /* go to the next level */ > > ext_debug("move to level %d (block %llu)\n", > > i + 1, idx_pblock(path[i].p_idx)); > > memset(path + i + 1, 0, sizeof(*path)); > > - path[i+1].p_bh = > > -
Re: [EXT4 set 2][PATCH 2/5] cleanups: Add extent sanity checks
On Thu, 12 Jul 2007 08:57:51 -0500 Dave Kleikamp <[EMAIL PROTECTED]> wrote: > On Thu, 2007-07-12 at 12:38 +0100, Andy Whitcroft wrote: > > Andrew Morton wrote: > > >> +if (ext4_ext_check_header(inode, > > >> ext_block_hdr(bh), > > >> +depth - i - 1)) > > >> { > > >> +err = -EIO; > > >> +break; > > >> +} > > >> +path[i+1].p_bh = bh; > > > > > > Really that should have been "i + 1". checkpatch misses this. It seems > > > to > > > be missing more stuff that it used to lately. > > > > This one is difficult. The rules up to now have been consistent spacing > > is required on both sides of mathematics operators. I personally like > > spaces always, but we do tend to use them without spaces too where the > > binding is effectivly part of the value -- the classic case is something > > like: > > > > pfn << MAX_ORDER-1 > > > > In allowing that sort of thing, we implictly allow the one you note > > above. We have tried to be overly annoying on these things, and so the > > check is consistancy, spaces both or neither. We could be stricter. > > I personally think stricter is better. An occasionally false-positive > isn't going to hurt anyone. (Well, maybe the checkpatch.pl maintainers > will get nagged.) It at least will cause the developer to look at the > line of code in question and make a conscious decision to leave it as it > is. I'm assuming that upstream maintainers use checkpatch.pl with some > constraint, and don't throw every patch that produces a warning back at > the submitter. > I'm in two minds. Missing-the-spaces is pretty damn common and is sometimes a reasonable way of saving quite a lot of horizontal space. I spose we could take it out again if it's causing problems. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [EXT4 set 2][PATCH 2/5] cleanups: Add extent sanity checks
On Thu, 2007-07-12 at 12:38 +0100, Andy Whitcroft wrote: > Andrew Morton wrote: > >> + if (ext4_ext_check_header(inode, ext_block_hdr(bh), > >> + depth - i - 1)) { > >> + err = -EIO; > >> + break; > >> + } > >> + path[i+1].p_bh = bh; > > > > Really that should have been "i + 1". checkpatch misses this. It seems to > > be missing more stuff that it used to lately. > > This one is difficult. The rules up to now have been consistent spacing > is required on both sides of mathematics operators. I personally like > spaces always, but we do tend to use them without spaces too where the > binding is effectivly part of the value -- the classic case is something > like: > > pfn << MAX_ORDER-1 > > In allowing that sort of thing, we implictly allow the one you note > above. We have tried to be overly annoying on these things, and so the > check is consistancy, spaces both or neither. We could be stricter. I personally think stricter is better. An occasionally false-positive isn't going to hurt anyone. (Well, maybe the checkpatch.pl maintainers will get nagged.) It at least will cause the developer to look at the line of code in question and make a conscious decision to leave it as it is. I'm assuming that upstream maintainers use checkpatch.pl with some constraint, and don't throw every patch that produces a warning back at the submitter. Shaggy -- David Kleikamp IBM Linux Technology Center - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [EXT4 set 2][PATCH 2/5] cleanups: Add extent sanity checks
Andrew Morton wrote: > On Sun, 01 Jul 2007 03:36:22 -0400 > Mingming Cao <[EMAIL PROTECTED]> wrote: > >> with the patch all headers are checked. the code should become >> more resistant to on-disk corruptions. needless BUG_ON() have >> been removed. please, review for inclusion. >> >> ... > >> @@ -269,6 +239,70 @@ >> return size; >> } >> >> +static inline int >> +ext4_ext_max_entries(struct inode *inode, int depth) > > Please remove the `inline'. > > `inline' is usually wrong. It is wrong here. If this function has a > single callsite (it has) then the compiler will inline it. If the function > later gains more callsites then the compiler will know (correctly) not to > inline it. > > We hope. If we find the compiler still inlines a function as large as this > one then we'd need to use noinline and complain at the gcc developers. > >> +{ >> +int max; >> + >> +if (depth == ext_depth(inode)) { >> +if (depth == 0) >> +max = ext4_ext_space_root(inode); >> +else >> +max = ext4_ext_space_root_idx(inode); >> +} else { >> +if (depth == 0) >> +max = ext4_ext_space_block(inode); >> +else >> +max = ext4_ext_space_block_idx(inode); >> +} >> + >> +return max; >> +} >> + >> +static int __ext4_ext_check_header(const char *function, struct inode >> *inode, >> +struct ext4_extent_header *eh, >> +int depth) >> +{ >> +const char *error_msg = NULL; > > Unneeded initialisation. > >> +int max = 0; >> + >> +if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) { >> +error_msg = "invalid magic"; >> +goto corrupted; >> +} >> +if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) { >> +error_msg = "unexpected eh_depth"; >> +goto corrupted; >> +} >> +if (unlikely(eh->eh_max == 0)) { >> +error_msg = "invalid eh_max"; >> +goto corrupted; >> +} >> +max = ext4_ext_max_entries(inode, depth); >> +if (unlikely(le16_to_cpu(eh->eh_max) > max)) { >> +error_msg = "too large eh_max"; >> +goto corrupted; >> +} >> +if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) { >> +error_msg = "invalid eh_entries"; >> +goto corrupted; >> +} >> +return 0; >> + >> +corrupted: >> +ext4_error(inode->i_sb, function, >> +"bad header in inode #%lu: %s - magic %x, " >> +"entries %u, max %u(%u), depth %u(%u)", >> +inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic), >> +le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max), >> +max, le16_to_cpu(eh->eh_depth), depth); >> + >> +return -EIO; >> +} >> + >> >> ... >> >> +i = depth = ext_depth(inode); >> > > Mulitple assignments like this are somewhat unpopular from the coding-style > POV. > > We have a local variable called `i' which is not used as a simple counter > and which has no explanatory comment. This is plain bad. Please find a > better name for this variable. Review the code for other such lack of > clarity - I'm sure there's more. > > >> -BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max)); >> -BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC); > > Yeah, this patch improves things a lot. > > How well-tested is it? Have any of these errors actually been triggered? > >> path[i].p_hdr = ext_block_hdr(path[i].p_bh); >> -if (ext4_ext_check_header(__FUNCTION__, inode, >> -path[i].p_hdr)) { >> -err = -EIO; >> -goto out; >> -} >> } >> >> -BUG_ON(le16_to_cpu(path[i].p_hdr->eh_entries) >> - > le16_to_cpu(path[i].p_hdr->eh_max)); >> -BUG_ON(path[i].p_hdr->eh_magic != EXT4_EXT_MAGIC); >> - >> if (!path[i].p_idx) { >> /* this level hasn't been touched yet */ >> path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); >> @@ -1873,17 +1890,24 @@ >> i, EXT_FIRST_INDEX(path[i].p_hdr), >> path[i].p_idx); >> if (ext4_ext_more_to_rm(path + i)) { >> +struct buffer_head *bh; >> /* go to the next level */ >> ext_debug("move to level %d (block %llu)\n", >>i + 1, idx_pblock(path[i].p_idx)); >> memset(path + i + 1, 0, sizeof(*path)); >> -path[i+1].p_bh = >> -sb_bread(sb, idx_pblock(path[i].p_idx)); >> -if (!path[i+1].p_bh) { >> +bh = sb_bread(sb, idx_pbloc
Re: [EXT4 set 2][PATCH 2/5] cleanups: Add extent sanity checks
On Sun, 01 Jul 2007 03:36:22 -0400 Mingming Cao <[EMAIL PROTECTED]> wrote: > with the patch all headers are checked. the code should become > more resistant to on-disk corruptions. needless BUG_ON() have > been removed. please, review for inclusion. > > ... > @@ -269,6 +239,70 @@ > return size; > } > > +static inline int > +ext4_ext_max_entries(struct inode *inode, int depth) Please remove the `inline'. `inline' is usually wrong. It is wrong here. If this function has a single callsite (it has) then the compiler will inline it. If the function later gains more callsites then the compiler will know (correctly) not to inline it. We hope. If we find the compiler still inlines a function as large as this one then we'd need to use noinline and complain at the gcc developers. > +{ > + int max; > + > + if (depth == ext_depth(inode)) { > + if (depth == 0) > + max = ext4_ext_space_root(inode); > + else > + max = ext4_ext_space_root_idx(inode); > + } else { > + if (depth == 0) > + max = ext4_ext_space_block(inode); > + else > + max = ext4_ext_space_block_idx(inode); > + } > + > + return max; > +} > + > +static int __ext4_ext_check_header(const char *function, struct inode *inode, > + struct ext4_extent_header *eh, > + int depth) > +{ > + const char *error_msg = NULL; Unneeded initialisation. > + int max = 0; > + > + if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) { > + error_msg = "invalid magic"; > + goto corrupted; > + } > + if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) { > + error_msg = "unexpected eh_depth"; > + goto corrupted; > + } > + if (unlikely(eh->eh_max == 0)) { > + error_msg = "invalid eh_max"; > + goto corrupted; > + } > + max = ext4_ext_max_entries(inode, depth); > + if (unlikely(le16_to_cpu(eh->eh_max) > max)) { > + error_msg = "too large eh_max"; > + goto corrupted; > + } > + if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) { > + error_msg = "invalid eh_entries"; > + goto corrupted; > + } > + return 0; > + > +corrupted: > + ext4_error(inode->i_sb, function, > + "bad header in inode #%lu: %s - magic %x, " > + "entries %u, max %u(%u), depth %u(%u)", > + inode->i_ino, error_msg, le16_to_cpu(eh->eh_magic), > + le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max), > + max, le16_to_cpu(eh->eh_depth), depth); > + > + return -EIO; > +} > + > > ... > > + i = depth = ext_depth(inode); > Mulitple assignments like this are somewhat unpopular from the coding-style POV. We have a local variable called `i' which is not used as a simple counter and which has no explanatory comment. This is plain bad. Please find a better name for this variable. Review the code for other such lack of clarity - I'm sure there's more. > - BUG_ON(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max)); > - BUG_ON(eh->eh_magic != EXT4_EXT_MAGIC); Yeah, this patch improves things a lot. How well-tested is it? Have any of these errors actually been triggered? > path[i].p_hdr = ext_block_hdr(path[i].p_bh); > - if (ext4_ext_check_header(__FUNCTION__, inode, > - path[i].p_hdr)) { > - err = -EIO; > - goto out; > - } > } > > - BUG_ON(le16_to_cpu(path[i].p_hdr->eh_entries) > -> le16_to_cpu(path[i].p_hdr->eh_max)); > - BUG_ON(path[i].p_hdr->eh_magic != EXT4_EXT_MAGIC); > - > if (!path[i].p_idx) { > /* this level hasn't been touched yet */ > path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr); > @@ -1873,17 +1890,24 @@ > i, EXT_FIRST_INDEX(path[i].p_hdr), > path[i].p_idx); > if (ext4_ext_more_to_rm(path + i)) { > + struct buffer_head *bh; > /* go to the next level */ > ext_debug("move to level %d (block %llu)\n", > i + 1, idx_pblock(path[i].p_idx)); > memset(path + i + 1, 0, sizeof(*path)); > - path[i+1].p_bh = > - sb_bread(sb, idx_pblock(path[i].p_idx)); > - if (!path[i+1].p_bh) { > + bh = sb_bread(sb, idx_pblock(path[i].p_idx)); > + if (!bh) { > /* should we reset i_size? */ >