tree: https://github.com/ceph/ceph-client.git wip-fscrypt head: cfa38f9d4169e3837d9a09b015492e9beb9bb8d6 commit: 1927572f5a964fa0e29bbd6fbe742c563ae66946 [38/54] ceph: handle fscrypt fields in cap messages from MDS config: arm-buildonly-randconfig-r003-20220331 (https://download.01.org/0day-ci/archive/20220401/202204010611.pgdnXxCx-lkp@xxxxxxxxx/config) compiler: arm-linux-gnueabi-gcc (GCC) 11.2.0 reproduce (this is a W=1 build): wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # https://github.com/ceph/ceph-client/commit/1927572f5a964fa0e29bbd6fbe742c563ae66946 git remote add ceph-client https://github.com/ceph/ceph-client.git git fetch --no-tags ceph-client wip-fscrypt git checkout 1927572f5a964fa0e29bbd6fbe742c563ae66946 # save the config file to linux build tree mkdir build_dir COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-11.2.0 make.cross O=build_dir ARCH=arm SHELL=/bin/bash fs/ If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot <lkp@xxxxxxxxx> All errors (new ones prefixed by >>): In file included from arch/arm/include/asm/bug.h:60, from include/linux/bug.h:5, from include/linux/thread_info.h:13, from include/asm-generic/preempt.h:5, from ./arch/arm/include/generated/asm/preempt.h:1, from include/linux/preempt.h:78, from include/linux/spinlock.h:55, from include/linux/wait.h:9, from include/linux/wait_bit.h:8, from include/linux/fs.h:6, from fs/ceph/caps.c:4: fs/ceph/caps.c: In function 'handle_cap_grant': >> fs/ceph/caps.c:3477:32: error: 'struct ceph_inode_info' has no member named 'fscrypt_auth_len' 3477 | WARN_ON_ONCE(ci->fscrypt_auth_len != extra_info->fscrypt_auth_len || | ^~ include/asm-generic/bug.h:166:32: note: in definition of macro 'WARN_ON' 166 | int __ret_warn_on = !!(condition); \ | ^~~~~~~~~ fs/ceph/caps.c:3477:17: note: in expansion of macro 'WARN_ON_ONCE' 3477 | WARN_ON_ONCE(ci->fscrypt_auth_len != extra_info->fscrypt_auth_len || | ^~~~~~~~~~~~ >> fs/ceph/caps.c:3478:39: error: 'struct ceph_inode_info' has no member named 'fscrypt_auth' 3478 | memcmp(ci->fscrypt_auth, extra_info->fscrypt_auth, | ^~ include/asm-generic/bug.h:166:32: note: in definition of macro 'WARN_ON' 166 | int __ret_warn_on = !!(condition); \ | ^~~~~~~~~ fs/ceph/caps.c:3477:17: note: in expansion of macro 'WARN_ON_ONCE' 3477 | WARN_ON_ONCE(ci->fscrypt_auth_len != extra_info->fscrypt_auth_len || | ^~~~~~~~~~~~ fs/ceph/caps.c:3479:40: error: 'struct ceph_inode_info' has no member named 'fscrypt_auth_len' 3479 | ci->fscrypt_auth_len)); | ^~ include/asm-generic/bug.h:166:32: note: in definition of macro 'WARN_ON' 166 | int __ret_warn_on = !!(condition); \ | ^~~~~~~~~ fs/ceph/caps.c:3477:17: note: in expansion of macro 'WARN_ON_ONCE' 3477 | WARN_ON_ONCE(ci->fscrypt_auth_len != extra_info->fscrypt_auth_len || | ^~~~~~~~~~~~ vim +3477 fs/ceph/caps.c 3371 3372 /* 3373 * Handle a cap GRANT message from the MDS. (Note that a GRANT may 3374 * actually be a revocation if it specifies a smaller cap set.) 3375 * 3376 * caller holds s_mutex and i_ceph_lock, we drop both. 3377 */ 3378 static void handle_cap_grant(struct inode *inode, 3379 struct ceph_mds_session *session, 3380 struct ceph_cap *cap, 3381 struct ceph_mds_caps *grant, 3382 struct ceph_buffer *xattr_buf, 3383 struct cap_extra_info *extra_info) 3384 __releases(ci->i_ceph_lock) 3385 __releases(session->s_mdsc->snap_rwsem) 3386 { 3387 struct ceph_inode_info *ci = ceph_inode(inode); 3388 int seq = le32_to_cpu(grant->seq); 3389 int newcaps = le32_to_cpu(grant->caps); 3390 int used, wanted, dirty; 3391 u64 size = le64_to_cpu(grant->size); 3392 u64 max_size = le64_to_cpu(grant->max_size); 3393 unsigned char check_caps = 0; 3394 bool was_stale = cap->cap_gen < atomic_read(&session->s_cap_gen); 3395 bool wake = false; 3396 bool writeback = false; 3397 bool queue_trunc = false; 3398 bool queue_invalidate = false; 3399 bool deleted_inode = false; 3400 bool fill_inline = false; 3401 3402 /* 3403 * If there is at least one crypto block then we'll trust fscrypt_file_size. 3404 * If the real length of the file is 0, then ignore it (it has probably been 3405 * truncated down to 0 by the MDS). 3406 */ 3407 if (IS_ENCRYPTED(inode) && size) 3408 size = extra_info->fscrypt_file_size; 3409 3410 dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n", 3411 inode, cap, session->s_mds, seq, ceph_cap_string(newcaps)); 3412 dout(" size %llu max_size %llu, i_size %llu\n", size, max_size, 3413 i_size_read(inode)); 3414 3415 3416 /* 3417 * If CACHE is being revoked, and we have no dirty buffers, 3418 * try to invalidate (once). (If there are dirty buffers, we 3419 * will invalidate _after_ writeback.) 3420 */ 3421 if (S_ISREG(inode->i_mode) && /* don't invalidate readdir cache */ 3422 ((cap->issued & ~newcaps) & CEPH_CAP_FILE_CACHE) && 3423 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0 && 3424 !(ci->i_wrbuffer_ref || ci->i_wb_ref)) { 3425 if (try_nonblocking_invalidate(inode)) { 3426 /* there were locked pages.. invalidate later 3427 in a separate thread. */ 3428 if (ci->i_rdcache_revoking != ci->i_rdcache_gen) { 3429 queue_invalidate = true; 3430 ci->i_rdcache_revoking = ci->i_rdcache_gen; 3431 } 3432 } 3433 } 3434 3435 if (was_stale) 3436 cap->issued = cap->implemented = CEPH_CAP_PIN; 3437 3438 /* 3439 * auth mds of the inode changed. we received the cap export message, 3440 * but still haven't received the cap import message. handle_cap_export 3441 * updated the new auth MDS' cap. 3442 * 3443 * "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing a message 3444 * that was sent before the cap import message. So don't remove caps. 3445 */ 3446 if (ceph_seq_cmp(seq, cap->seq) <= 0) { 3447 WARN_ON(cap != ci->i_auth_cap); 3448 WARN_ON(cap->cap_id != le64_to_cpu(grant->cap_id)); 3449 seq = cap->seq; 3450 newcaps |= cap->issued; 3451 } 3452 3453 /* side effects now are allowed */ 3454 cap->cap_gen = atomic_read(&session->s_cap_gen); 3455 cap->seq = seq; 3456 3457 __check_cap_issue(ci, cap, newcaps); 3458 3459 inode_set_max_iversion_raw(inode, extra_info->change_attr); 3460 3461 if ((newcaps & CEPH_CAP_AUTH_SHARED) && 3462 (extra_info->issued & CEPH_CAP_AUTH_EXCL) == 0) { 3463 umode_t mode = le32_to_cpu(grant->mode); 3464 3465 if (inode_wrong_type(inode, mode)) 3466 pr_warn_once("inode type changed! (ino %llx.%llx is 0%o, mds says 0%o)\n", 3467 ceph_vinop(inode), inode->i_mode, mode); 3468 else 3469 inode->i_mode = mode; 3470 inode->i_uid = make_kuid(&init_user_ns, le32_to_cpu(grant->uid)); 3471 inode->i_gid = make_kgid(&init_user_ns, le32_to_cpu(grant->gid)); 3472 ci->i_btime = extra_info->btime; 3473 dout("%p mode 0%o uid.gid %d.%d\n", inode, inode->i_mode, 3474 from_kuid(&init_user_ns, inode->i_uid), 3475 from_kgid(&init_user_ns, inode->i_gid)); 3476 > 3477 WARN_ON_ONCE(ci->fscrypt_auth_len != extra_info->fscrypt_auth_len || > 3478 memcmp(ci->fscrypt_auth, extra_info->fscrypt_auth, 3479 ci->fscrypt_auth_len)); 3480 } 3481 3482 if ((newcaps & CEPH_CAP_LINK_SHARED) && 3483 (extra_info->issued & CEPH_CAP_LINK_EXCL) == 0) { 3484 set_nlink(inode, le32_to_cpu(grant->nlink)); 3485 if (inode->i_nlink == 0) 3486 deleted_inode = true; 3487 } 3488 3489 if ((extra_info->issued & CEPH_CAP_XATTR_EXCL) == 0 && 3490 grant->xattr_len) { 3491 int len = le32_to_cpu(grant->xattr_len); 3492 u64 version = le64_to_cpu(grant->xattr_version); 3493 3494 if (version > ci->i_xattrs.version) { 3495 dout(" got new xattrs v%llu on %p len %d\n", 3496 version, inode, len); 3497 if (ci->i_xattrs.blob) 3498 ceph_buffer_put(ci->i_xattrs.blob); 3499 ci->i_xattrs.blob = ceph_buffer_get(xattr_buf); 3500 ci->i_xattrs.version = version; 3501 ceph_forget_all_cached_acls(inode); 3502 ceph_security_invalidate_secctx(inode); 3503 } 3504 } 3505 3506 if (newcaps & CEPH_CAP_ANY_RD) { 3507 struct timespec64 mtime, atime, ctime; 3508 /* ctime/mtime/atime? */ 3509 ceph_decode_timespec64(&mtime, &grant->mtime); 3510 ceph_decode_timespec64(&atime, &grant->atime); 3511 ceph_decode_timespec64(&ctime, &grant->ctime); 3512 ceph_fill_file_time(inode, extra_info->issued, 3513 le32_to_cpu(grant->time_warp_seq), 3514 &ctime, &mtime, &atime); 3515 } 3516 3517 if ((newcaps & CEPH_CAP_FILE_SHARED) && extra_info->dirstat_valid) { 3518 ci->i_files = extra_info->nfiles; 3519 ci->i_subdirs = extra_info->nsubdirs; 3520 } 3521 3522 if (newcaps & (CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR)) { 3523 /* file layout may have changed */ 3524 s64 old_pool = ci->i_layout.pool_id; 3525 struct ceph_string *old_ns; 3526 3527 ceph_file_layout_from_legacy(&ci->i_layout, &grant->layout); 3528 old_ns = rcu_dereference_protected(ci->i_layout.pool_ns, 3529 lockdep_is_held(&ci->i_ceph_lock)); 3530 rcu_assign_pointer(ci->i_layout.pool_ns, extra_info->pool_ns); 3531 3532 if (ci->i_layout.pool_id != old_pool || 3533 extra_info->pool_ns != old_ns) 3534 ci->i_ceph_flags &= ~CEPH_I_POOL_PERM; 3535 3536 extra_info->pool_ns = old_ns; 3537 3538 /* size/truncate_seq? */ 3539 queue_trunc = ceph_fill_file_size(inode, extra_info->issued, 3540 le32_to_cpu(grant->truncate_seq), 3541 le64_to_cpu(grant->truncate_size), 3542 size); 3543 } 3544 3545 if (ci->i_auth_cap == cap && (newcaps & CEPH_CAP_ANY_FILE_WR)) { 3546 if (max_size != ci->i_max_size) { 3547 dout("max_size %lld -> %llu\n", 3548 ci->i_max_size, max_size); 3549 ci->i_max_size = max_size; 3550 if (max_size >= ci->i_wanted_max_size) { 3551 ci->i_wanted_max_size = 0; /* reset */ 3552 ci->i_requested_max_size = 0; 3553 } 3554 wake = true; 3555 } 3556 } 3557 3558 /* check cap bits */ 3559 wanted = __ceph_caps_wanted(ci); 3560 used = __ceph_caps_used(ci); 3561 dirty = __ceph_caps_dirty(ci); 3562 dout(" my wanted = %s, used = %s, dirty %s\n", 3563 ceph_cap_string(wanted), 3564 ceph_cap_string(used), 3565 ceph_cap_string(dirty)); 3566 3567 if ((was_stale || le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) && 3568 (wanted & ~(cap->mds_wanted | newcaps))) { 3569 /* 3570 * If mds is importing cap, prior cap messages that update 3571 * 'wanted' may get dropped by mds (migrate seq mismatch). 3572 * 3573 * We don't send cap message to update 'wanted' if what we 3574 * want are already issued. If mds revokes caps, cap message 3575 * that releases caps also tells mds what we want. But if 3576 * caps got revoked by mds forcedly (session stale). We may 3577 * haven't told mds what we want. 3578 */ 3579 check_caps = 1; 3580 } 3581 3582 /* revocation, grant, or no-op? */ 3583 if (cap->issued & ~newcaps) { 3584 int revoking = cap->issued & ~newcaps; 3585 3586 dout("revocation: %s -> %s (revoking %s)\n", 3587 ceph_cap_string(cap->issued), 3588 ceph_cap_string(newcaps), 3589 ceph_cap_string(revoking)); 3590 if (S_ISREG(inode->i_mode) && 3591 (revoking & used & CEPH_CAP_FILE_BUFFER)) 3592 writeback = true; /* initiate writeback; will delay ack */ 3593 else if (queue_invalidate && 3594 revoking == CEPH_CAP_FILE_CACHE && 3595 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0) 3596 ; /* do nothing yet, invalidation will be queued */ 3597 else if (cap == ci->i_auth_cap) 3598 check_caps = 1; /* check auth cap only */ 3599 else 3600 check_caps = 2; /* check all caps */ 3601 cap->issued = newcaps; 3602 cap->implemented |= newcaps; 3603 } else if (cap->issued == newcaps) { 3604 dout("caps unchanged: %s -> %s\n", 3605 ceph_cap_string(cap->issued), ceph_cap_string(newcaps)); 3606 } else { 3607 dout("grant: %s -> %s\n", ceph_cap_string(cap->issued), 3608 ceph_cap_string(newcaps)); 3609 /* non-auth MDS is revoking the newly grant caps ? */ 3610 if (cap == ci->i_auth_cap && 3611 __ceph_caps_revoking_other(ci, cap, newcaps)) 3612 check_caps = 2; 3613 3614 cap->issued = newcaps; 3615 cap->implemented |= newcaps; /* add bits only, to 3616 * avoid stepping on a 3617 * pending revocation */ 3618 wake = true; 3619 } 3620 BUG_ON(cap->issued & ~cap->implemented); 3621 3622 if (extra_info->inline_version > 0 && 3623 extra_info->inline_version >= ci->i_inline_version) { 3624 ci->i_inline_version = extra_info->inline_version; 3625 if (ci->i_inline_version != CEPH_INLINE_NONE && 3626 (newcaps & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO))) 3627 fill_inline = true; 3628 } 3629 3630 if (ci->i_auth_cap == cap && 3631 le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) { 3632 if (newcaps & ~extra_info->issued) 3633 wake = true; 3634 3635 if (ci->i_requested_max_size > max_size || 3636 !(le32_to_cpu(grant->wanted) & CEPH_CAP_ANY_FILE_WR)) { 3637 /* re-request max_size if necessary */ 3638 ci->i_requested_max_size = 0; 3639 wake = true; 3640 } 3641 3642 ceph_kick_flushing_inode_caps(session, ci); 3643 spin_unlock(&ci->i_ceph_lock); 3644 up_read(&session->s_mdsc->snap_rwsem); 3645 } else { 3646 spin_unlock(&ci->i_ceph_lock); 3647 } 3648 3649 if (fill_inline) 3650 ceph_fill_inline_data(inode, NULL, extra_info->inline_data, 3651 extra_info->inline_len); 3652 3653 if (queue_trunc) 3654 ceph_queue_vmtruncate(inode); 3655 3656 if (writeback) 3657 /* 3658 * queue inode for writeback: we can't actually call 3659 * filemap_write_and_wait, etc. from message handler 3660 * context. 3661 */ 3662 ceph_queue_writeback(inode); 3663 if (queue_invalidate) 3664 ceph_queue_invalidate(inode); 3665 if (deleted_inode) 3666 invalidate_aliases(inode); 3667 if (wake) 3668 wake_up_all(&ci->i_cap_wq); 3669 3670 mutex_unlock(&session->s_mutex); 3671 if (check_caps == 1) 3672 ceph_check_caps(ci, CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL, 3673 session); 3674 else if (check_caps == 2) 3675 ceph_check_caps(ci, CHECK_CAPS_NOINVAL, session); 3676 } 3677 -- 0-DAY CI Kernel Test Service https://01.org/lkp
