From 089ceecc1ea4a69ed8bcc5c7c7b96ce487e26b33 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Sun, 5 Jul 2009 22:17:31 -0400 Subject: ext4: mark several more functions in mballoc.c as noinline Ted noticed a stack-deep callchain through writepages->ext4_mb_regular_allocator->ext4_mb_init_cache->submit_bh ... With all the static functions in mballoc.c, gcc helpfully inlines for us, and we get something like this: ext4_mb_regular_allocator (232 bytes stack) ext4_mb_init_cache (232 bytes stack) submit_bh (starts 464 deeper) the 2 ext4 functions here get several others inlined; by telling gcc not to inline them, we can save stack space for when we head off into submit_bh land and associated block layer callchains. The following noinlined functions are only called once, so this won't impact any other callchains: ext4_mb_regular_allocator (104) (was 232) ext4_mb_find_by_goal (56) (noinlined) ext4_mb_init_group (24) (noinlined) ext4_mb_init_cache (136) (was 232) ext4_mb_generate_buddy (88) (noinlined) ext4_mb_generate_from_pa (40) (noinlined) submit_bh ext4_mb_simple_scan_group (24) (noinlined) ext4_mb_scan_aligned (56) (noinlined) ext4_mb_complex_scan_group (40) (noinlined) ext4_mb_try_best_found (24) (noinlined) now when we head off into submit_bh() we're only 264 bytes deeper in stack than when we entered ext4_mb_regular_allocator() (vs. 464 bytes before). Every 200 bytes helps. :) Signed-off-by: Eric Sandeen Signed-off-by: "Theodore Ts'o" --- fs/ext4/mballoc.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) (limited to 'fs/ext4/mballoc.c') diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index 519a0a686d94..4a45efabb203 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -657,7 +657,8 @@ static void ext4_mb_mark_free_simple(struct super_block *sb, } } -static void ext4_mb_generate_buddy(struct super_block *sb, +static noinline_for_stack +void ext4_mb_generate_buddy(struct super_block *sb, void *buddy, void *bitmap, ext4_group_t group) { struct ext4_group_info *grp = ext4_get_group_info(sb, group); @@ -1480,7 +1481,8 @@ static void ext4_mb_measure_extent(struct ext4_allocation_context *ac, ext4_mb_check_limits(ac, e4b, 0); } -static int ext4_mb_try_best_found(struct ext4_allocation_context *ac, +static noinline_for_stack +int ext4_mb_try_best_found(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { struct ext4_free_extent ex = ac->ac_b_ex; @@ -1507,7 +1509,8 @@ static int ext4_mb_try_best_found(struct ext4_allocation_context *ac, return 0; } -static int ext4_mb_find_by_goal(struct ext4_allocation_context *ac, +static noinline_for_stack +int ext4_mb_find_by_goal(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { ext4_group_t group = ac->ac_g_ex.fe_group; @@ -1566,7 +1569,8 @@ static int ext4_mb_find_by_goal(struct ext4_allocation_context *ac, * The routine scans buddy structures (not bitmap!) from given order * to max order and tries to find big enough chunk to satisfy the req */ -static void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac, +static noinline_for_stack +void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; @@ -1609,7 +1613,8 @@ static void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac, * In order to optimize scanning, caller must pass number of * free blocks in the group, so the routine can know upper limit. */ -static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, +static noinline_for_stack +void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; @@ -1668,7 +1673,8 @@ static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, * we try to find stripe-aligned chunks for stripe-size requests * XXX should do so at least for multiples of stripe size as well */ -static void ext4_mb_scan_aligned(struct ext4_allocation_context *ac, +static noinline_for_stack +void ext4_mb_scan_aligned(struct ext4_allocation_context *ac, struct ext4_buddy *e4b) { struct super_block *sb = ac->ac_sb; @@ -1831,7 +1837,8 @@ void ext4_mb_put_buddy_cache_lock(struct super_block *sb, } -static int ext4_mb_init_group(struct super_block *sb, ext4_group_t group) +static noinline_for_stack +int ext4_mb_init_group(struct super_block *sb, ext4_group_t group) { int ret; @@ -3457,7 +3464,8 @@ static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, * used in in-core bitmap. buddy must be generated from this bitmap * Need to be called with ext4 group lock held */ -static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, +static noinline_for_stack +void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, ext4_group_t group) { struct ext4_group_info *grp = ext4_get_group_info(sb, group); -- cgit v1.2.3 From 3e03f9ca6a2599db1823bb0ea24e0845219a0e69 Mon Sep 17 00:00:00 2001 From: Jesper Dangaard Brouer Date: Sun, 5 Jul 2009 22:29:27 -0400 Subject: ext4: Use rcu_barrier() on module unload. The ext4 module uses rcu_call() thus it should use rcu_barrier()on module unload. The kmem cache ext4_pspace_cachep is sometimes free'ed using call_rcu() callbacks. Thus, we must wait for completion of call_rcu() before doing kmem_cache_destroy(). Signed-off-by: Jesper Dangaard Brouer Signed-off-by: "Theodore Ts'o" --- fs/ext4/mballoc.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/ext4/mballoc.c') diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index 4a45efabb203..2fcaf286f1de 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -2909,7 +2909,11 @@ int __init init_ext4_mballoc(void) void exit_ext4_mballoc(void) { - /* XXX: synchronize_rcu(); */ + /* + * Wait for completion of call_rcu()'s on ext4_pspace_cachep + * before destroying the slab cache. + */ + rcu_barrier(); kmem_cache_destroy(ext4_pspace_cachep); kmem_cache_destroy(ext4_ac_cachep); kmem_cache_destroy(ext4_free_ext_cachep); -- cgit v1.2.3 From 833576b362e15c38be3bfe43942cda693e56287c Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Mon, 13 Jul 2009 09:45:52 -0400 Subject: ext4: Fix ext4_mb_initialize_context() to initialize all fields Pavel Roskin pointed out that kmemcheck indicated that ext4_mb_store_history() was accessing uninitialized values of ac->ac_tail and ac->ac_buddy leading to garbage in the mballoc history. Fix this by initializing the entire structure to all zeros first. Also, two fields were getting doubly initialized by the caller of ext4_mb_initialize_context, so remove them for efficiency's sake. Signed-off-by: "Theodore Ts'o" --- fs/ext4/mballoc.c | 20 ++------------------ 1 file changed, 2 insertions(+), 18 deletions(-) (limited to 'fs/ext4/mballoc.c') diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index 2fcaf286f1de..cd258463e2a9 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -4227,14 +4227,9 @@ ext4_mb_initialize_context(struct ext4_allocation_context *ac, ext4_get_group_no_and_offset(sb, goal, &group, &block); /* set up allocation goals */ + memset(ac, 0, sizeof(struct ext4_allocation_context)); ac->ac_b_ex.fe_logical = ar->logical; - ac->ac_b_ex.fe_group = 0; - ac->ac_b_ex.fe_start = 0; - ac->ac_b_ex.fe_len = 0; ac->ac_status = AC_STATUS_CONTINUE; - ac->ac_groups_scanned = 0; - ac->ac_ex_scanned = 0; - ac->ac_found = 0; ac->ac_sb = sb; ac->ac_inode = ar->inode; ac->ac_o_ex.fe_logical = ar->logical; @@ -4245,15 +4240,7 @@ ext4_mb_initialize_context(struct ext4_allocation_context *ac, ac->ac_g_ex.fe_group = group; ac->ac_g_ex.fe_start = block; ac->ac_g_ex.fe_len = len; - ac->ac_f_ex.fe_len = 0; ac->ac_flags = ar->flags; - ac->ac_2order = 0; - ac->ac_criteria = 0; - ac->ac_pa = NULL; - ac->ac_bitmap_page = NULL; - ac->ac_buddy_page = NULL; - ac->alloc_semp = NULL; - ac->ac_lg = NULL; /* we have to define context: we'll we work with a file or * locality group. this is a policy, actually */ @@ -4521,10 +4508,7 @@ ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle, } ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS); - if (ac) { - ac->ac_sb = sb; - ac->ac_inode = ar->inode; - } else { + if (!ac) { ar->len = 0; *errp = -ENOMEM; goto out1; -- cgit v1.2.3