From f5cd27ec714628b522af8c588d3907671404385e Mon Sep 17 00:00:00 2001 From: Gabriel Shahrouzi Date: Thu, 3 Apr 2025 21:28:20 -0400 Subject: [PATCH 1/8] bcachefs: Fix escape sequence in prt_printf Remove backslash before format specifier. Ensure correct output. Signed-off-by: Gabriel Shahrouzi Signed-off-by: Kent Overstreet --- fs/bcachefs/data_update.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c index de02ebf847ec..b211c97238ab 100644 --- a/fs/bcachefs/data_update.c +++ b/fs/bcachefs/data_update.c @@ -607,7 +607,7 @@ void bch2_data_update_inflight_to_text(struct printbuf *out, struct data_update prt_newline(out); printbuf_indent_add(out, 2); bch2_data_update_opts_to_text(out, m->op.c, &m->op.opts, &m->data_opts); - prt_printf(out, "read_done:\t\%u\n", m->read_done); + prt_printf(out, "read_done:\t%u\n", m->read_done); bch2_write_op_to_text(out, &m->op); printbuf_indent_sub(out, 2); } From afc5444e4d862c805798016a37cf301f5f30e1e6 Mon Sep 17 00:00:00 2001 From: Gabriel Shahrouzi Date: Thu, 3 Apr 2025 21:28:21 -0400 Subject: [PATCH 2/8] bcachefs: Fix type for parameter in journal_advance_devs_to_next_bucket Replace u64 with __le64 to match the expected parameter type. Ensure consistency both in function calls and within the function itself. Signed-off-by: Gabriel Shahrouzi Signed-off-by: Kent Overstreet --- fs/bcachefs/journal_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c index 1b7961f4f609..2a54ac79189b 100644 --- a/fs/bcachefs/journal_io.c +++ b/fs/bcachefs/journal_io.c @@ -1460,7 +1460,7 @@ int bch2_journal_read(struct bch_fs *c, static void journal_advance_devs_to_next_bucket(struct journal *j, struct dev_alloc_list *devs, - unsigned sectors, u64 seq) + unsigned sectors, __le64 seq) { struct bch_fs *c = container_of(j, struct bch_fs, journal); From 4a22a7332341f49e60b804811fee7edf87909e9e Mon Sep 17 00:00:00 2001 From: Gabriel Shahrouzi Date: Thu, 3 Apr 2025 21:28:22 -0400 Subject: [PATCH 3/8] bcachefs: Use cpu_to_le16 for dirent lengths Prevent incorrect byte ordering for big-endian systems. Signed-off-by: Gabriel Shahrouzi Signed-off-by: Kent Overstreet --- fs/bcachefs/dirent.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/bcachefs/dirent.c b/fs/bcachefs/dirent.c index bf53a029f356..8488a7578115 100644 --- a/fs/bcachefs/dirent.c +++ b/fs/bcachefs/dirent.c @@ -287,8 +287,8 @@ static void dirent_init_casefolded_name(struct bkey_i_dirent *dirent, EBUG_ON(!dirent->v.d_casefold); EBUG_ON(!cf_name->len); - dirent->v.d_cf_name_block.d_name_len = name->len; - dirent->v.d_cf_name_block.d_cf_name_len = cf_name->len; + dirent->v.d_cf_name_block.d_name_len = cpu_to_le16(name->len); + dirent->v.d_cf_name_block.d_cf_name_len = cpu_to_le16(cf_name->len); memcpy(&dirent->v.d_cf_name_block.d_names[0], name->name, name->len); memcpy(&dirent->v.d_cf_name_block.d_names[name->len], cf_name->name, cf_name->len); memset(&dirent->v.d_cf_name_block.d_names[name->len + cf_name->len], 0, From 34b47e3d73a21ef992905746cdb044ce02d3b29a Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sat, 5 Apr 2025 12:26:43 -0400 Subject: [PATCH 4/8] bcachefs: Fix UAF in bchfs_read() Commit 3ba0240a8789 fixed a bug in the read retry path in __bch2_read(), and changed bchfs_read() to match - to avoid a landmine if bch2_read_extent() ever starts returning transaction restarts. But that was incorrect, because bchfs_read() doesn't use a separate stack allocated bvec_iter, it uses the one in the rbio being submitted. Add a comment explaining the issue, and revert the buggy change. Fixes: 3ba0240a8789 ("bcachefs: Fix silent short reads in data read retry path") Reported-by: syzbot+2deb10b8dc9aae6fab67@syzkaller.appspotmail.com Signed-off-by: Kent Overstreet --- fs/bcachefs/fs-io-buffered.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) diff --git a/fs/bcachefs/fs-io-buffered.c b/fs/bcachefs/fs-io-buffered.c index 19d4599918dc..e3a75dcca60c 100644 --- a/fs/bcachefs/fs-io-buffered.c +++ b/fs/bcachefs/fs-io-buffered.c @@ -225,11 +225,26 @@ static void bchfs_read(struct btree_trans *trans, bch2_read_extent(trans, rbio, iter.pos, data_btree, k, offset_into_extent, flags); - swap(rbio->bio.bi_iter.bi_size, bytes); + /* + * Careful there's a landmine here if bch2_read_extent() ever + * starts returning transaction restarts here. + * + * We've changed rbio->bi_iter.bi_size to be "bytes we can read + * from this extent" with the swap call, and we restore it + * below. That restore needs to come before checking for + * errors. + * + * But unlike __bch2_read(), we use the rbio bvec iter, not one + * on the stack, so we can't do the restore right after the + * bch2_read_extent() call: we don't own that iterator anymore + * if BCH_READ_last_fragment is set, since we may have submitted + * that rbio instead of cloning it. + */ if (flags & BCH_READ_last_fragment) break; + swap(rbio->bio.bi_iter.bi_size, bytes); bio_advance(&rbio->bio, bytes); err: if (ret && From 1ec94a9f6dd870a4d9c799d26e440c8f6b6268d2 Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Sun, 6 Apr 2025 13:38:11 -0400 Subject: [PATCH 5/8] bcachefs: Fix duplicate "ro,read_only" in opts at startup Signed-off-by: Kent Overstreet --- fs/bcachefs/super.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index a58edde43bee..16ebea1214af 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -1002,12 +1002,6 @@ static void print_mount_opts(struct bch_fs *c) prt_str(&p, "starting version "); bch2_version_to_text(&p, c->sb.version); - if (c->opts.read_only) { - prt_str(&p, " opts="); - first = false; - prt_printf(&p, "ro"); - } - for (i = 0; i < bch2_opts_nr; i++) { const struct bch_option *opt = &bch2_opt_table[i]; u64 v = bch2_opt_get_by_id(&c->opts, i); From 4bf4b5046de0ef7f9dc50f3a9ef8a6dcda178a6d Mon Sep 17 00:00:00 2001 From: Eric Biggers Date: Tue, 1 Apr 2025 21:33:33 -0700 Subject: [PATCH 6/8] bcachefs: use library APIs for ChaCha20 and Poly1305 Just use the ChaCha20 and Poly1305 libraries instead of the clunky crypto API. This is much simpler. It is also slightly faster, since the libraries provide more direct access to the same architecture-optimized ChaCha20 and Poly1305 code. I've tested that existing encrypted bcachefs filesystems can be continue to be accessed with this patch applied. Reviewed-by: Ard Biesheuvel Signed-off-by: Eric Biggers Signed-off-by: Kent Overstreet --- fs/bcachefs/Kconfig | 5 +- fs/bcachefs/bcachefs.h | 4 +- fs/bcachefs/btree_node_scan.c | 2 +- fs/bcachefs/checksum.c | 245 ++++++++-------------------------- fs/bcachefs/checksum.h | 3 +- fs/bcachefs/io_read.c | 3 +- fs/bcachefs/super.c | 3 - 7 files changed, 64 insertions(+), 201 deletions(-) diff --git a/fs/bcachefs/Kconfig b/fs/bcachefs/Kconfig index bf1c94e51dd0..508c0cfdf7c9 100644 --- a/fs/bcachefs/Kconfig +++ b/fs/bcachefs/Kconfig @@ -15,10 +15,9 @@ config BCACHEFS_FS select ZLIB_INFLATE select ZSTD_COMPRESS select ZSTD_DECOMPRESS - select CRYPTO select CRYPTO_LIB_SHA256 - select CRYPTO_CHACHA20 - select CRYPTO_POLY1305 + select CRYPTO_LIB_CHACHA + select CRYPTO_LIB_POLY1305 select KEYS select RAID6_PQ select XOR_BLOCKS diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index 5d9f208a1bb7..5cb0fc384ac0 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -981,8 +981,8 @@ struct bch_fs { mempool_t compress_workspace[BCH_COMPRESSION_OPT_NR]; size_t zstd_workspace_size; - struct crypto_sync_skcipher *chacha20; - struct crypto_shash *poly1305; + struct bch_key chacha20_key; + bool chacha20_key_set; atomic64_t key_version; diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c index 8c9fdb7263fe..f520fbcd0455 100644 --- a/fs/bcachefs/btree_node_scan.c +++ b/fs/bcachefs/btree_node_scan.c @@ -183,7 +183,7 @@ static void try_read_btree_node(struct find_btree_nodes *f, struct bch_dev *ca, return; if (bch2_csum_type_is_encryption(BSET_CSUM_TYPE(&bn->keys))) { - if (!c->chacha20) + if (!c->chacha20_key_set) return; struct nonce nonce = btree_nonce(&bn->keys, 0); diff --git a/fs/bcachefs/checksum.c b/fs/bcachefs/checksum.c index 3726689093e3..d0a34a097b80 100644 --- a/fs/bcachefs/checksum.c +++ b/fs/bcachefs/checksum.c @@ -7,17 +7,12 @@ #include "super-io.h" #include -#include #include #include #include #include -#include -#include #include -#include #include -#include #include /* @@ -96,116 +91,40 @@ static void bch2_checksum_update(struct bch2_checksum_state *state, const void * } } -static inline int do_encrypt_sg(struct crypto_sync_skcipher *tfm, - struct nonce nonce, - struct scatterlist *sg, size_t len) +static void bch2_chacha20_init(u32 state[CHACHA_STATE_WORDS], + const struct bch_key *key, struct nonce nonce) { - SYNC_SKCIPHER_REQUEST_ON_STACK(req, tfm); + u32 key_words[CHACHA_KEY_SIZE / sizeof(u32)]; - skcipher_request_set_sync_tfm(req, tfm); - skcipher_request_set_callback(req, 0, NULL, NULL); - skcipher_request_set_crypt(req, sg, sg, len, nonce.d); + BUILD_BUG_ON(sizeof(key_words) != sizeof(*key)); + memcpy(key_words, key, sizeof(key_words)); + le32_to_cpu_array(key_words, ARRAY_SIZE(key_words)); - int ret = crypto_skcipher_encrypt(req); - if (ret) - pr_err("got error %i from crypto_skcipher_encrypt()", ret); + BUILD_BUG_ON(sizeof(nonce) != CHACHA_IV_SIZE); + chacha_init(state, key_words, (const u8 *)nonce.d); - return ret; + memzero_explicit(key_words, sizeof(key_words)); } -static inline int do_encrypt(struct crypto_sync_skcipher *tfm, - struct nonce nonce, - void *buf, size_t len) +static void bch2_chacha20(const struct bch_key *key, struct nonce nonce, + void *data, size_t len) { - if (!is_vmalloc_addr(buf)) { - struct scatterlist sg = {}; + u32 state[CHACHA_STATE_WORDS]; - sg_mark_end(&sg); - sg_set_page(&sg, virt_to_page(buf), len, offset_in_page(buf)); - return do_encrypt_sg(tfm, nonce, &sg, len); - } else { - DARRAY_PREALLOCATED(struct scatterlist, 4) sgl; - size_t sgl_len = 0; - int ret; - - darray_init(&sgl); - - while (len) { - unsigned offset = offset_in_page(buf); - struct scatterlist sg = { - .page_link = (unsigned long) vmalloc_to_page(buf), - .offset = offset, - .length = min(len, PAGE_SIZE - offset), - }; - - if (darray_push(&sgl, sg)) { - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(tfm, nonce, sgl.data, sgl_len); - if (ret) - goto err; - - nonce = nonce_add(nonce, sgl_len); - sgl_len = 0; - sgl.nr = 0; - BUG_ON(darray_push(&sgl, sg)); - } - - buf += sg.length; - len -= sg.length; - sgl_len += sg.length; - } - - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(tfm, nonce, sgl.data, sgl_len); -err: - darray_exit(&sgl); - return ret; - } + bch2_chacha20_init(state, key, nonce); + chacha20_crypt(state, data, data, len); + memzero_explicit(state, sizeof(state)); } -int bch2_chacha_encrypt_key(struct bch_key *key, struct nonce nonce, - void *buf, size_t len) +static void bch2_poly1305_init(struct poly1305_desc_ctx *desc, + struct bch_fs *c, struct nonce nonce) { - struct crypto_sync_skcipher *chacha20 = - crypto_alloc_sync_skcipher("chacha20", 0, 0); - int ret; - - ret = PTR_ERR_OR_ZERO(chacha20); - if (ret) { - pr_err("error requesting chacha20 cipher: %s", bch2_err_str(ret)); - return ret; - } - - ret = crypto_skcipher_setkey(&chacha20->base, - (void *) key, sizeof(*key)); - if (ret) { - pr_err("error from crypto_skcipher_setkey(): %s", bch2_err_str(ret)); - goto err; - } - - ret = do_encrypt(chacha20, nonce, buf, len); -err: - crypto_free_sync_skcipher(chacha20); - return ret; -} - -static int gen_poly_key(struct bch_fs *c, struct shash_desc *desc, - struct nonce nonce) -{ - u8 key[POLY1305_KEY_SIZE]; - int ret; + u8 key[POLY1305_KEY_SIZE] = { 0 }; nonce.d[3] ^= BCH_NONCE_POLY; - memset(key, 0, sizeof(key)); - ret = do_encrypt(c->chacha20, nonce, key, sizeof(key)); - if (ret) - return ret; - - desc->tfm = c->poly1305; - crypto_shash_init(desc); - crypto_shash_update(desc, key, sizeof(key)); - return 0; + bch2_chacha20(&c->chacha20_key, nonce, key, sizeof(key)); + poly1305_init(desc, key); } struct bch_csum bch2_checksum(struct bch_fs *c, unsigned type, @@ -230,14 +149,13 @@ struct bch_csum bch2_checksum(struct bch_fs *c, unsigned type, case BCH_CSUM_chacha20_poly1305_80: case BCH_CSUM_chacha20_poly1305_128: { - SHASH_DESC_ON_STACK(desc, c->poly1305); + struct poly1305_desc_ctx dctx; u8 digest[POLY1305_DIGEST_SIZE]; struct bch_csum ret = { 0 }; - gen_poly_key(c, desc, nonce); - - crypto_shash_update(desc, data, len); - crypto_shash_final(desc, digest); + bch2_poly1305_init(&dctx, c, nonce); + poly1305_update(&dctx, data, len); + poly1305_final(&dctx, digest); memcpy(&ret, digest, bch_crc_bytes[type]); return ret; @@ -253,11 +171,12 @@ int bch2_encrypt(struct bch_fs *c, unsigned type, if (!bch2_csum_type_is_encryption(type)) return 0; - if (bch2_fs_inconsistent_on(!c->chacha20, + if (bch2_fs_inconsistent_on(!c->chacha20_key_set, c, "attempting to encrypt without encryption key")) return -BCH_ERR_no_encryption_key; - return do_encrypt(c->chacha20, nonce, data, len); + bch2_chacha20(&c->chacha20_key, nonce, data, len); + return 0; } static struct bch_csum __bch2_checksum_bio(struct bch_fs *c, unsigned type, @@ -296,26 +215,26 @@ static struct bch_csum __bch2_checksum_bio(struct bch_fs *c, unsigned type, case BCH_CSUM_chacha20_poly1305_80: case BCH_CSUM_chacha20_poly1305_128: { - SHASH_DESC_ON_STACK(desc, c->poly1305); + struct poly1305_desc_ctx dctx; u8 digest[POLY1305_DIGEST_SIZE]; struct bch_csum ret = { 0 }; - gen_poly_key(c, desc, nonce); + bch2_poly1305_init(&dctx, c, nonce); #ifdef CONFIG_HIGHMEM __bio_for_each_segment(bv, bio, *iter, *iter) { void *p = kmap_local_page(bv.bv_page) + bv.bv_offset; - crypto_shash_update(desc, p, bv.bv_len); + poly1305_update(&dctx, p, bv.bv_len); kunmap_local(p); } #else __bio_for_each_bvec(bv, bio, *iter, *iter) - crypto_shash_update(desc, + poly1305_update(&dctx, page_address(bv.bv_page) + bv.bv_offset, bv.bv_len); #endif - crypto_shash_final(desc, digest); + poly1305_final(&dctx, digest); memcpy(&ret, digest, bch_crc_bytes[type]); return ret; @@ -338,43 +257,33 @@ int __bch2_encrypt_bio(struct bch_fs *c, unsigned type, { struct bio_vec bv; struct bvec_iter iter; - DARRAY_PREALLOCATED(struct scatterlist, 4) sgl; - size_t sgl_len = 0; + u32 chacha_state[CHACHA_STATE_WORDS]; int ret = 0; - if (bch2_fs_inconsistent_on(!c->chacha20, + if (bch2_fs_inconsistent_on(!c->chacha20_key_set, c, "attempting to encrypt without encryption key")) return -BCH_ERR_no_encryption_key; - darray_init(&sgl); + bch2_chacha20_init(chacha_state, &c->chacha20_key, nonce); bio_for_each_segment(bv, bio, iter) { - struct scatterlist sg = { - .page_link = (unsigned long) bv.bv_page, - .offset = bv.bv_offset, - .length = bv.bv_len, - }; + void *p; - if (darray_push(&sgl, sg)) { - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(c->chacha20, nonce, sgl.data, sgl_len); - if (ret) - goto err; - - nonce = nonce_add(nonce, sgl_len); - sgl_len = 0; - sgl.nr = 0; - - BUG_ON(darray_push(&sgl, sg)); + /* + * chacha_crypt() assumes that the length is a multiple of + * CHACHA_BLOCK_SIZE on any non-final call. + */ + if (!IS_ALIGNED(bv.bv_len, CHACHA_BLOCK_SIZE)) { + bch_err_ratelimited(c, "bio not aligned for encryption"); + ret = -EIO; + break; } - sgl_len += sg.length; + p = bvec_kmap_local(&bv); + chacha20_crypt(chacha_state, p, p, bv.bv_len); + kunmap_local(p); } - - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(c->chacha20, nonce, sgl.data, sgl_len); -err: - darray_exit(&sgl); + memzero_explicit(chacha_state, sizeof(chacha_state)); return ret; } @@ -650,10 +559,7 @@ int bch2_decrypt_sb_key(struct bch_fs *c, } /* decrypt real key: */ - ret = bch2_chacha_encrypt_key(&user_key, bch2_sb_key_nonce(c), - &sb_key, sizeof(sb_key)); - if (ret) - goto err; + bch2_chacha20(&user_key, bch2_sb_key_nonce(c), &sb_key, sizeof(sb_key)); if (bch2_key_is_encrypted(&sb_key)) { bch_err(c, "incorrect encryption key"); @@ -668,31 +574,6 @@ int bch2_decrypt_sb_key(struct bch_fs *c, return ret; } -static int bch2_alloc_ciphers(struct bch_fs *c) -{ - if (c->chacha20) - return 0; - - struct crypto_sync_skcipher *chacha20 = crypto_alloc_sync_skcipher("chacha20", 0, 0); - int ret = PTR_ERR_OR_ZERO(chacha20); - if (ret) { - bch_err(c, "error requesting chacha20 module: %s", bch2_err_str(ret)); - return ret; - } - - struct crypto_shash *poly1305 = crypto_alloc_shash("poly1305", 0, 0); - ret = PTR_ERR_OR_ZERO(poly1305); - if (ret) { - bch_err(c, "error requesting poly1305 module: %s", bch2_err_str(ret)); - crypto_free_sync_skcipher(chacha20); - return ret; - } - - c->chacha20 = chacha20; - c->poly1305 = poly1305; - return 0; -} - #if 0 /* @@ -797,35 +678,21 @@ int bch2_enable_encryption(struct bch_fs *c, bool keyed) void bch2_fs_encryption_exit(struct bch_fs *c) { - if (c->poly1305) - crypto_free_shash(c->poly1305); - if (c->chacha20) - crypto_free_sync_skcipher(c->chacha20); + memzero_explicit(&c->chacha20_key, sizeof(c->chacha20_key)); } int bch2_fs_encryption_init(struct bch_fs *c) { struct bch_sb_field_crypt *crypt; - struct bch_key key; - int ret = 0; + int ret; crypt = bch2_sb_field_get(c->disk_sb.sb, crypt); if (!crypt) - goto out; + return 0; - ret = bch2_alloc_ciphers(c); + ret = bch2_decrypt_sb_key(c, crypt, &c->chacha20_key); if (ret) - goto out; - - ret = bch2_decrypt_sb_key(c, crypt, &key); - if (ret) - goto out; - - ret = crypto_skcipher_setkey(&c->chacha20->base, - (void *) &key.key, sizeof(key.key)); - if (ret) - goto out; -out: - memzero_explicit(&key, sizeof(key)); - return ret; + return ret; + c->chacha20_key_set = true; + return 0; } diff --git a/fs/bcachefs/checksum.h b/fs/bcachefs/checksum.h index 4ac251c8fcd8..1310782d3ae9 100644 --- a/fs/bcachefs/checksum.h +++ b/fs/bcachefs/checksum.h @@ -69,7 +69,6 @@ static inline void bch2_csum_err_msg(struct printbuf *out, bch2_csum_to_text(out, type, expected); } -int bch2_chacha_encrypt_key(struct bch_key *, struct nonce, void *, size_t); int bch2_request_key(struct bch_sb *, struct bch_key *); #ifndef __KERNEL__ int bch2_revoke_key(struct bch_sb *); @@ -156,7 +155,7 @@ static inline bool bch2_checksum_type_valid(const struct bch_fs *c, if (type >= BCH_CSUM_NR) return false; - if (bch2_csum_type_is_encryption(type) && !c->chacha20) + if (bch2_csum_type_is_encryption(type) && !c->chacha20_key_set) return false; return true; diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c index 417bb0c7bbfa..fd627c8d1053 100644 --- a/fs/bcachefs/io_read.c +++ b/fs/bcachefs/io_read.c @@ -977,7 +977,8 @@ int __bch2_read_extent(struct btree_trans *trans, struct bch_read_bio *orig, goto err; } - if (unlikely(bch2_csum_type_is_encryption(pick.crc.csum_type)) && !c->chacha20) { + if (unlikely(bch2_csum_type_is_encryption(pick.crc.csum_type)) && + !c->chacha20_key_set) { struct printbuf buf = PRINTBUF; bch2_read_err_msg_trans(trans, &buf, orig, read_pos); prt_printf(&buf, "attempting to read encrypted data without encryption key\n "); diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index 16ebea1214af..956f51d563b4 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -70,13 +70,10 @@ #include #include #include -#include MODULE_LICENSE("GPL"); MODULE_AUTHOR("Kent Overstreet "); MODULE_DESCRIPTION("bcachefs filesystem"); -MODULE_SOFTDEP("pre: chacha20"); -MODULE_SOFTDEP("pre: poly1305"); MODULE_SOFTDEP("pre: xxhash"); const char * const bch2_fs_flag_strs[] = { From c92896ffb7fc0c49ca031d9e98e76008597bbab0 Mon Sep 17 00:00:00 2001 From: Eric Biggers Date: Wed, 2 Apr 2025 09:23:38 -0700 Subject: [PATCH 7/8] bcachefs: Remove unnecessary softdep on xxhash As with the other algorithms, bcachefs does not access xxhash through the crypto API. So there is no need to use a module softdep to ensure that it is loaded. Signed-off-by: Eric Biggers Signed-off-by: Kent Overstreet --- fs/bcachefs/super.c | 1 - 1 file changed, 1 deletion(-) diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index 956f51d563b4..b79e80a435e0 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -74,7 +74,6 @@ MODULE_LICENSE("GPL"); MODULE_AUTHOR("Kent Overstreet "); MODULE_DESCRIPTION("bcachefs filesystem"); -MODULE_SOFTDEP("pre: xxhash"); const char * const bch2_fs_flag_strs[] = { #define x(n) #n, From 55fd97fbc4744a43fb2d134908e481266cf33bda Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Wed, 26 Mar 2025 11:44:30 -0400 Subject: [PATCH 8/8] bcachefs: Use sort_nonatomic() instead of sort() Fixes "task out to lunch" warnings during recovery on large machines with lots of dirty data in the journal. Signed-off-by: Kent Overstreet --- fs/bcachefs/btree_journal_iter.c | 5 ++--- fs/bcachefs/btree_node_scan.c | 4 ++-- fs/bcachefs/btree_write_buffer.c | 8 ++++---- fs/bcachefs/recovery.c | 6 +++--- 4 files changed, 11 insertions(+), 12 deletions(-) diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c index d1ad1a7613c9..7d6c971db23c 100644 --- a/fs/bcachefs/btree_journal_iter.c +++ b/fs/bcachefs/btree_journal_iter.c @@ -644,8 +644,6 @@ void bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans, */ static int journal_sort_key_cmp(const void *_l, const void *_r) { - cond_resched(); - const struct journal_key *l = _l; const struct journal_key *r = _r; @@ -689,7 +687,8 @@ void bch2_journal_keys_put(struct bch_fs *c) static void __journal_keys_sort(struct journal_keys *keys) { - sort(keys->data, keys->nr, sizeof(keys->data[0]), journal_sort_key_cmp, NULL); + sort_nonatomic(keys->data, keys->nr, sizeof(keys->data[0]), + journal_sort_key_cmp, NULL); cond_resched(); diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c index f520fbcd0455..86acf037590c 100644 --- a/fs/bcachefs/btree_node_scan.c +++ b/fs/bcachefs/btree_node_scan.c @@ -398,7 +398,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c) bch2_print_string_as_lines(KERN_INFO, buf.buf); } - sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL); + sort_nonatomic(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL); dst = 0; darray_for_each(f->nodes, i) { @@ -418,7 +418,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c) } f->nodes.nr = dst; - sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); + sort_nonatomic(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); if (0 && c->opts.verbose) { printbuf_reset(&buf); diff --git a/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c index adbe576ec77e..0941fb2c026d 100644 --- a/fs/bcachefs/btree_write_buffer.c +++ b/fs/bcachefs/btree_write_buffer.c @@ -428,10 +428,10 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans) */ trace_and_count(c, write_buffer_flush_slowpath, trans, slowpath, wb->flushing.keys.nr); - sort(wb->flushing.keys.data, - wb->flushing.keys.nr, - sizeof(wb->flushing.keys.data[0]), - wb_key_seq_cmp, NULL); + sort_nonatomic(wb->flushing.keys.data, + wb->flushing.keys.nr, + sizeof(wb->flushing.keys.data[0]), + wb_key_seq_cmp, NULL); darray_for_each(wb->flushing.keys, i) { if (!i->journal_seq) diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 79fd18a5a07c..d2b07f602da9 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -389,9 +389,9 @@ int bch2_journal_replay(struct bch_fs *c) * Now, replay any remaining keys in the order in which they appear in * the journal, unpinning those journal entries as we go: */ - sort(keys_sorted.data, keys_sorted.nr, - sizeof(keys_sorted.data[0]), - journal_sort_seq_cmp, NULL); + sort_nonatomic(keys_sorted.data, keys_sorted.nr, + sizeof(keys_sorted.data[0]), + journal_sort_seq_cmp, NULL); darray_for_each(keys_sorted, kp) { cond_resched();