From 8c61291fd8500e3b35c7ec0c781b273d8cc96cde Mon Sep 17 00:00:00 2001 From: Zhaoyang Huang Date: Fri, 7 Jun 2024 10:31:16 +0800 Subject: mm: fix incorrect vbq reference in purge_fragmented_block xa_for_each() in _vm_unmap_aliases() loops through all vbs. However, since commit 062eacf57ad9 ("mm: vmalloc: remove a global vmap_blocks xarray") the vb from xarray may not be on the corresponding CPU vmap_block_queue. Consequently, purge_fragmented_block() might use the wrong vbq->lock to protect the free list, leading to vbq->free breakage. Incorrect lock protection can exhaust all vmalloc space as follows: CPU0 CPU1 +--------------------------------------------+ | +--------------------+ +-----+ | +--> | |---->| |------+ | CPU1:vbq free_list | | vb1 | +--- | |<----| |<-----+ | +--------------------+ +-----+ | +--------------------------------------------+ _vm_unmap_aliases() vb_alloc() new_vmap_block() xa_for_each(&vbq->vmap_blocks, idx, vb) --> vb in CPU1:vbq->freelist purge_fragmented_block(vb) spin_lock(&vbq->lock) spin_lock(&vbq->lock) --> use CPU0:vbq->lock --> use CPU1:vbq->lock list_del_rcu(&vb->free_list) list_add_tail_rcu(&vb->free_list, &vbq->free) __list_del(vb->prev, vb->next) next->prev = prev +--------------------+ | | | CPU1:vbq free_list | +---| |<--+ | +--------------------+ | +----------------------------+ __list_add(new, head->prev, head) +--------------------------------------------+ | +--------------------+ +-----+ | +--> | |---->| |------+ | CPU1:vbq free_list | | vb2 | +--- | |<----| |<-----+ | +--------------------+ +-----+ | +--------------------------------------------+ prev->next = next +--------------------------------------------+ |----------------------------+ | | +--------------------+ | +-----+ | +--> | |--+ | |------+ | CPU1:vbq free_list | | vb2 | +--- | |<----| |<-----+ | +--------------------+ +-----+ | +--------------------------------------------+ Here’s a list breakdown. All vbs, which were to be added to ‘prev’, cannot be used by list_for_each_entry_rcu(vb, &vbq->free, free_list) in vb_alloc(). Thus, vmalloc space is exhausted. This issue affects both erofs and f2fs, the stacktrace is as follows: erofs: [] __switch_to+0x174 [] __schedule+0x624 [] schedule+0x7c [] schedule_preempt_disabled+0x24 [] __mutex_lock+0x374 [] __mutex_lock_slowpath+0x14 [] mutex_lock+0x24 [] reclaim_and_purge_vmap_areas+0x44 [] alloc_vmap_area+0x2e0 [] vm_map_ram+0x1b0 [] z_erofs_lz4_decompress+0x278 [] z_erofs_decompress_queue+0x650 [] z_erofs_runqueue+0x7f4 [] z_erofs_read_folio+0x104 [] filemap_read_folio+0x6c [] filemap_fault+0x300 [] __do_fault+0xc8 [] handle_mm_fault+0xb38 [] do_page_fault+0x288 [] do_translation_fault[jt]+0x40 [] do_mem_abort+0x58 [] el0_ia+0x70 [] el0t_64_sync_handler[jt]+0xb0 [] ret_to_user[jt]+0x0 f2fs: [] __switch_to+0x174 [] __schedule+0x624 [] schedule+0x7c [] schedule_preempt_disabled+0x24 [] __mutex_lock+0x374 [] __mutex_lock_slowpath+0x14 [] mutex_lock+0x24 [] reclaim_and_purge_vmap_areas+0x44 [] alloc_vmap_area+0x2e0 [] vm_map_ram+0x1b0 [] f2fs_prepare_decomp_mem+0x144 [] f2fs_alloc_dic+0x264 [] f2fs_read_multi_pages+0x428 [] f2fs_mpage_readpages+0x314 [] f2fs_readahead+0x50 [] read_pages+0x80 [] page_cache_ra_unbounded+0x1a0 [] page_cache_ra_order+0x274 [] do_sync_mmap_readahead+0x11c [] filemap_fault+0x1a0 [] f2fs_filemap_fault+0x28 [] __do_fault+0xc8 [] handle_mm_fault+0xb38 [] do_page_fault+0x288 [] do_translation_fault[jt]+0x40 [] do_mem_abort+0x58 [] el0_ia+0x70 [] el0t_64_sync_handler[jt]+0xb0 [] ret_to_user[jt]+0x0 To fix this, introducee cpu within vmap_block to record which this vb belongs to. Link: https://lkml.kernel.org/r/20240614021352.1822225-1-zhaoyang.huang@unisoc.com Link: https://lkml.kernel.org/r/20240607023116.1720640-1-zhaoyang.huang@unisoc.com Fixes: fc1e0d980037 ("mm/vmalloc: prevent stale TLBs in fully utilized blocks") Signed-off-by: Zhaoyang Huang Suggested-by: Hailong.Liu Reviewed-by: Uladzislau Rezki (Sony) Cc: Baoquan He Cc: Christoph Hellwig Cc: Lorenzo Stoakes Cc: Thomas Gleixner Cc: Signed-off-by: Andrew Morton --- mm/vmalloc.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) (limited to 'mm') diff --git a/mm/vmalloc.c b/mm/vmalloc.c index 45e1506d58c3..d0cbdd7c1e5b 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -2498,6 +2498,7 @@ struct vmap_block { struct list_head free_list; struct rcu_head rcu_head; struct list_head purge; + unsigned int cpu; }; /* Queue of free and dirty vmap blocks, for allocation and flushing purposes */ @@ -2625,8 +2626,15 @@ static void *new_vmap_block(unsigned int order, gfp_t gfp_mask) free_vmap_area(va); return ERR_PTR(err); } - - vbq = raw_cpu_ptr(&vmap_block_queue); + /* + * list_add_tail_rcu could happened in another core + * rather than vb->cpu due to task migration, which + * is safe as list_add_tail_rcu will ensure the list's + * integrity together with list_for_each_rcu from read + * side. + */ + vb->cpu = raw_smp_processor_id(); + vbq = per_cpu_ptr(&vmap_block_queue, vb->cpu); spin_lock(&vbq->lock); list_add_tail_rcu(&vb->free_list, &vbq->free); spin_unlock(&vbq->lock); @@ -2654,9 +2662,10 @@ static void free_vmap_block(struct vmap_block *vb) } static bool purge_fragmented_block(struct vmap_block *vb, - struct vmap_block_queue *vbq, struct list_head *purge_list, - bool force_purge) + struct list_head *purge_list, bool force_purge) { + struct vmap_block_queue *vbq = &per_cpu(vmap_block_queue, vb->cpu); + if (vb->free + vb->dirty != VMAP_BBMAP_BITS || vb->dirty == VMAP_BBMAP_BITS) return false; @@ -2704,7 +2713,7 @@ static void purge_fragmented_blocks(int cpu) continue; spin_lock(&vb->lock); - purge_fragmented_block(vb, vbq, &purge, true); + purge_fragmented_block(vb, &purge, true); spin_unlock(&vb->lock); } rcu_read_unlock(); @@ -2841,7 +2850,7 @@ static void _vm_unmap_aliases(unsigned long start, unsigned long end, int flush) * not purgeable, check whether there is dirty * space to be flushed. */ - if (!purge_fragmented_block(vb, vbq, &purge_list, false) && + if (!purge_fragmented_block(vb, &purge_list, false) && vb->dirty_max && vb->dirty != VMAP_BBMAP_BITS) { unsigned long va_start = vb->va->va_start; unsigned long s, e; -- cgit v1.2.3