From: "Hailong.Liu" <hailong.liu@xxxxxxxx> The function 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: [<ffffffd4ffb93ad4>] __switch_to+0x174 [<ffffffd4ffb942f0>] __schedule+0x624 [<ffffffd4ffb946f4>] schedule+0x7c [<ffffffd4ffb947cc>] schedule_preempt_disabled+0x24 [<ffffffd4ffb962ec>] __mutex_lock+0x374 [<ffffffd4ffb95998>] __mutex_lock_slowpath+0x14 [<ffffffd4ffb95954>] mutex_lock+0x24 [<ffffffd4fef2900c>] reclaim_and_purge_vmap_areas+0x44 [<ffffffd4fef25908>] alloc_vmap_area+0x2e0 [<ffffffd4fef24ea0>] vm_map_ram+0x1b0 [<ffffffd4ff1b46f4>] z_erofs_lz4_decompress+0x278 [<ffffffd4ff1b8ac4>] z_erofs_decompress_queue+0x650 [<ffffffd4ff1b8328>] z_erofs_runqueue+0x7f4 [<ffffffd4ff1b66a8>] z_erofs_read_folio+0x104 [<ffffffd4feeb6fec>] filemap_read_folio+0x6c [<ffffffd4feeb68c4>] filemap_fault+0x300 [<ffffffd4fef0ecac>] __do_fault+0xc8 [<ffffffd4fef0c908>] handle_mm_fault+0xb38 [<ffffffd4ffb9f008>] do_page_fault+0x288 [<ffffffd4ffb9ed64>] do_translation_fault[jt]+0x40 [<ffffffd4fec39c78>] do_mem_abort+0x58 [<ffffffd4ffb8c3e4>] el0_ia+0x70 [<ffffffd4ffb8c260>] el0t_64_sync_handler[jt]+0xb0 [<ffffffd4fec11588>] ret_to_user[jt]+0x0 f2fs: [<ffffffd4ffb93ad4>] __switch_to+0x174 [<ffffffd4ffb942f0>] __schedule+0x624 [<ffffffd4ffb946f4>] schedule+0x7c [<ffffffd4ffb947cc>] schedule_preempt_disabled+0x24 [<ffffffd4ffb962ec>] __mutex_lock+0x374 [<ffffffd4ffb95998>] __mutex_lock_slowpath+0x14 [<ffffffd4ffb95954>] mutex_lock+0x24 [<ffffffd4fef2900c>] reclaim_and_purge_vmap_areas+0x44 [<ffffffd4fef25908>] alloc_vmap_area+0x2e0 [<ffffffd4fef24ea0>] vm_map_ram+0x1b0 [<ffffffd4ff1a3b60>] f2fs_prepare_decomp_mem+0x144 [<ffffffd4ff1a6c24>] f2fs_alloc_dic+0x264 [<ffffffd4ff175468>] f2fs_read_multi_pages+0x428 [<ffffffd4ff17b46c>] f2fs_mpage_readpages+0x314 [<ffffffd4ff1785c4>] f2fs_readahead+0x50 [<ffffffd4feec3384>] read_pages+0x80 [<ffffffd4feec32c0>] page_cache_ra_unbounded+0x1a0 [<ffffffd4feec39e8>] page_cache_ra_order+0x274 [<ffffffd4feeb6cec>] do_sync_mmap_readahead+0x11c [<ffffffd4feeb6764>] filemap_fault+0x1a0 [<ffffffd4ff1423bc>] f2fs_filemap_fault+0x28 [<ffffffd4fef0ecac>] __do_fault+0xc8 [<ffffffd4fef0c908>] handle_mm_fault+0xb38 [<ffffffd4ffb9f008>] do_page_fault+0x288 [<ffffffd4ffb9ed64>] do_translation_fault[jt]+0x40 [<ffffffd4fec39c78>] do_mem_abort+0x58 [<ffffffd4ffb8c3e4>] el0_ia+0x70 [<ffffffd4ffb8c260>] el0t_64_sync_handler[jt]+0xb0 [<ffffffd4fec11588>] ret_to_user[jt]+0x0 To fix this, introduce vbq_lock in vmap_block. Cc: <stable@xxxxxxxxxxxxxxx> Fixes: fc1e0d980037 ("mm/vmalloc: prevent stale TLBs in fully utilized blocks") Suggested-by: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx> Signed-off-by: Hailong.Liu <hailong.liu@xxxxxxxx> --- mm/vmalloc.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/mm/vmalloc.c b/mm/vmalloc.c index 125427cbdb87..f4cfdf3fd925 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -2467,6 +2467,7 @@ struct vmap_block_queue { struct vmap_block { spinlock_t lock; + spinlock_t *vbq_lock; struct vmap_area *va; unsigned long free, dirty; DECLARE_BITMAP(used_map, VMAP_BBMAP_BITS); @@ -2603,6 +2604,7 @@ static void *new_vmap_block(unsigned int order, gfp_t gfp_mask) } vbq = raw_cpu_ptr(&vmap_block_queue); + vb->vbq_lock = &vbq->lock; spin_lock(&vbq->lock); list_add_tail_rcu(&vb->free_list, &vbq->free); spin_unlock(&vbq->lock); @@ -2630,7 +2632,7 @@ 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, + struct list_head *purge_list, bool force_purge) { if (vb->free + vb->dirty != VMAP_BBMAP_BITS || @@ -2647,9 +2649,9 @@ static bool purge_fragmented_block(struct vmap_block *vb, WRITE_ONCE(vb->dirty, VMAP_BBMAP_BITS); vb->dirty_min = 0; vb->dirty_max = VMAP_BBMAP_BITS; - spin_lock(&vbq->lock); + spin_lock(vb->vbq_lock); list_del_rcu(&vb->free_list); - spin_unlock(&vbq->lock); + spin_unlock(vb->vbq_lock); list_add_tail(&vb->purge, purge_list); return true; } @@ -2680,7 +2682,7 @@ 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(); @@ -2817,7 +2819,7 @@ 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; --- Changes since v2 [2]: - simply revert the patch has other effects, per Zhaoyang Changes since v1 [1]: - add runtime effect in commit msg, per Andrew. BTW, 1. use xa_for_each to iterate all vb, we need a mapping from vb to vbq. Zhaoyang use cpuid to get vbq as follows: https://lore.kernel.org/all/20240531030520.1615833-1-zhaoyang.huang@xxxxxxxxxx/ IMO, why not just store the address of the lock, which might waste a few more bytes but would look clearer. Baoquan and Hillf save directly to the queue correspoding to va, - vbq = raw_cpu_ptr(&vmap_block_queue); + vbq = container_of(xa, struct vmap_block_queue, vmap_blocks); spin_lock(&vbq->lock); , /* * We should probably have a fallback mechanism to allocate virtual memory * out of partially filled vmap blocks. However vmap block sizing should be @@ -2626,7 +2634,7 @@ static void *new_vmap_block(unsigned int order, gfp_t gfp_mask) return ERR_PTR(err); } - vbq = raw_cpu_ptr(&vmap_block_queue); + vbq = addr_to_vbq(va->va_start); If in this case we actually don't need the percpu variable at all, because each address directly to the correspoding index through hash function. Does anyone have a btter suggestion? https://lore.kernel.org/all/ZlqIp9V1Jknm7axa@MiWiFi-R3L-srv/ [1] https://lore.kernel.org/all/20240530093108.4512-1-hailong.liu@xxxxxxxx/ [2] https://lore.kernel.org/all/20240531024820.5507-1-hailong.liu@xxxxxxxx/ -- 2.34.1