dlm: fix recovery pending middle conversion
Alexander Aring
maintainerPATCH 1/7 dlm: fix recovery pending middle conversion
A workload involving PR <-> CW conversions and triggering recovery can end in a so named "conversion deadlock" situation that is signaled by "dlm: WARN: pending deadlock 1e node 0 2 1bf21" in the kernel log.
Under normal circumstances such conversion deadlocks are solved immediately, in this case recovery created such scenario that was not solved immediately. This scenario that two locks ending up on the convertqueue with the conversion PR -> CW. In normal circumstances one of the conversion will be rejected with -DEADLK as CW cannot be granted when one lock is helding still PR. Usually one of those conversion will immediately rejected and the rejected conversion need to convert to a compatible lock mode. If such situation is created on the convertqueue we don't solve such conversion in the expected way by the user.
The situation is created by recovery when a pending middle conversion will be recovered and signaled by:
receive_rcom_lock_args 2e middle convert gr 3 rq 2 remote 2 1e
In this case recovery will remove waiting for the pending message and force the lock being on the convertqueue without checking if there is another incompatible conversion going on like PR -> CW which was the case as the mentioned above "WARN pending deadlock ..." occurs.
This state is difficult to reproduce as it is requires a pending PR -> CW conversion, however we automated a test scenario that fences randomly on PR -> CW conversion and we was able to hit it.
The proposed change in this patch changes to not "force" putting the pending middle conversion on the convertqueue and just handle it like every other message to resend it later to the new lock master. To using the existing convert functionality we will immediately reject such conversion if a incompatible mode on the convertqueue is detected.
Long run with the automated randomly middle conversion test showed so far we don't run into a "WARN: pending deadlock ..." situation again.
--- fs/dlm/lock.c | 19 +------------------ 1 file changed, 1 insertion(+), 18 deletions(-)
fs/dlm/lock.c +1 -18
@@ -5014,25 +5014,8 @@ void dlm_receive_buffer(const union dlm_packet *p, int nodeid)static void recover_convert_waiter (struct dlm_ls *ls,struct dlm_lkb *lkb,struct dlm_message *ms_local) { -if (middle_conversion (lkb)) { -log_rinfo (ls,"%s %x middle convert in progress" , __func__, - lkb->lkb_id); - -/* We sent this lock to the new master. The new master will -* tell us when it's granted. We no longer need a reply, so -* use a fake reply to put the lkb into the right state. -*/ -hold_lkb (lkb); -memset (ms_local,0 ,sizeof (struct dlm_message )); - ms_local->m_type =cpu_to_le32 (DLM_MSG_CONVERT_REPLY ); - ms_local->m_result =cpu_to_le32 (to_dlm_errno (-EINPROGRESS )); - ms_local->m_header.h_nodeid =cpu_to_le32 (lkb->lkb_nodeid); -_receive_convert_reply (lkb, ms_local, true); -unhold_lkb (lkb); - - }else if (lkb->lkb_rqmode >= lkb->lkb_grmode) { +if (middle_conversion (lkb) || lkb->lkb_rqmode >= lkb->lkb_grmode)set_bit (DLM_IFL_RESEND_BIT , &lkb->lkb_iflags); - }/* lkb->lkb_rqmode < lkb->lkb_grmode shouldn't happen since down conversions are async; there's no reply from the remote master */
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 2/7 dlm: validate length in dlm_search_rsb_tree
From: Ezrak1e <[email protected]>
The len parameter in dlm_dump_rsb_name() is not validated and comes from network messages. When it exceeds DLM_RESNAME_MAXLEN, it can cause out-of-bounds write in dlm_search_rsb_tree().
Add length validation to prevent potential buffer overflow.
--- fs/dlm/lock.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
fs/dlm/lock.c +2 -1
@@ -626,7 +626,8 @@ int dlm_search_rsb_tree(struct rhashtable *rhash, const void *name, int len, struct dlm_rsb **r_ret) {char key[DLM_RESNAME_MAXLEN ] = {}; - +if (len >DLM_RESNAME_MAXLEN ) +return -EINVAL ;memcpy (key, name, len); *r_ret =rhashtable_lookup_fast (rhash, &key, dlm_rhash_rsb_params);if (*r_ret)
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 3/7 fs/dlm: use list_add_tail() instead of open-coding list insertion
In reply to Alexander AringFrom: Shaurya Rane <[email protected]>
Replace the manual list pointer manipulation in add_ordered_member() with the standard list_add_tail() helper. The original code explicitly updated ->prev and ->next pointers to insert @newlist before @tmp, which is exactly what list_add_tail(newlist, tmp) provides.
Using the list macro improves readability, removes a source of potential pointer bugs, and satisfies the existing FIXME requesting conversion to the list helpers. No functional change in the ordering logic for DLM members.
--- fs/dlm/member.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-)
fs/dlm/member.c +1 -5
@@ -299,11 +299,7 @@ static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new)if (!memb)list_add_tail (newlist, head);else { -/* FIXME: can use list macro here */ - newlist->prev = tmp->prev; - newlist->next = tmp; - tmp->prev->next = newlist; - tmp->prev = newlist; +list_add_tail (newlist, tmp); } }
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 4/7 dlm: Constify struct configfs_item_operations and configfs_group_operations
In reply to Alexander AringFrom: Christophe JAILLET <[email protected]>
'struct configfs_item_operations' and 'configfs_group_operations' are not modified in this driver.
Constifying these structures moves some data to a read-only section, so increases overall security, especially when the structure holds some function pointers.
On a x86_64, with allmodconfig, as an example:
Before:
======
text data bss dec hex filename
29436 12952 384 42772 a714 fs/dlm/config.o
After:
=====
text data bss dec hex filename
30076 12312 384 42772 a714 fs/dlm/config.o
--- fs/dlm/config.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
fs/dlm/config.c +8 -8
@@ -324,39 +324,39 @@ struct dlm_member_gone {struct list_head list;/* space->members_gone */ }; -static struct configfs_group_operations clusters_ops = { +static const struct configfs_group_operations clusters_ops = { .make_group = make_cluster, .drop_item = drop_cluster, }; -static struct configfs_item_operations cluster_ops = { +static const struct configfs_item_operations cluster_ops = { .release = release_cluster, }; -static struct configfs_group_operations spaces_ops = { +static const struct configfs_group_operations spaces_ops = { .make_group = make_space, .drop_item = drop_space, }; -static struct configfs_item_operations space_ops = { +static const struct configfs_item_operations space_ops = { .release = release_space, }; -static struct configfs_group_operations comms_ops = { +static const struct configfs_group_operations comms_ops = { .make_item = make_comm, .drop_item = drop_comm, }; -static struct configfs_item_operations comm_ops = { +static const struct configfs_item_operations comm_ops = { .release = release_comm, }; -static struct configfs_group_operations nodes_ops = { +static const struct configfs_group_operations nodes_ops = { .make_item = make_node, .drop_item = drop_node, }; -static struct configfs_item_operations node_ops = { +static const struct configfs_item_operations node_ops = { .release = release_node, };
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 5/7 fs/dlm/dir: remove unuse variable count_match
In reply to Alexander AringFrom: Alex Shi <[email protected]>
The variable was never used after introduced. Better to comment it if we want to keep the info.
fs/dlm/dir.c:65:26: error: variable 'count_match' set but not used [-Werror,-Wunused-but-set-variable]
65 | unsigned int count = 0, count_match = 0, count_bad = 0, count_add = 0;
| ^
1 error generated.
--- fs/dlm/dir.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
fs/dlm/dir.c +3 -3
@@ -62,7 +62,7 @@ int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq)char *b, *last_name =NULL ;int error = -ENOMEM , last_len, nodeid, result;uint16_t namelen; -unsigned int count =0 , count_match =0 , count_bad =0 , count_add =0 ; +unsigned int count =0 , count_bad =0 , count_add =0 ;log_rinfo (ls,"dlm_recover_directory" );
@@ -157,12 +157,12 @@ int dlm_recover_directory(struct dlm_ls *ls, uint64_t seq) }/* The name was found in rsbtbl, and the -* master nodeid matches memb->nodeid. */ +* master nodeid matches memb->nodeid. if (result == DLM_LU_MATCH && nodeid == memb->nodeid) { count_match++; - } +}*/ /* The name was not found in rsbtbl and was * added with memb->nodeid as the master. */
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 6/7 dlm: use bool for coniditonal expressions
In reply to Alexander AringThis patch changes the return value of functions instead of using int and their corresponding value 1/0 for true/false. Instead we use the predefined true/false definitions.
--- fs/dlm/dlm_internal.h | 2 +- fs/dlm/lock.c | 151 +++++++++++++++++++++--------------------- fs/dlm/lowcomms.c | 16 ++--- fs/dlm/member.c | 18 ++--- fs/dlm/member.h | 6 +- fs/dlm/requestqueue.c | 12 ++-- fs/dlm/user.c | 16 ++--- fs/dlm/user.h | 2 +- 8 files changed, 111 insertions(+), 112 deletions(-)
fs/dlm/dlm_internal.h +1 -1
@@ -381,7 +381,7 @@ static inline void rsb_clear_flag(struct dlm_rsb *r, enum rsb_flags flag)__clear_bit (flag, &r->res_flags); } -static inline int rsb_flag (struct dlm_rsb *r,enum rsb_flags flag) +static inline bool rsb_flag (struct dlm_rsb *r,enum rsb_flags flag) {return test_bit (flag, &r->res_flags); }
fs/dlm/lock.c +75 -76
423 lines not on this page · raw message
fs/dlm/lowcomms.c +8 -8
@@ -331,32 +331,32 @@ static struct connection *nodeid2con(int nodeid, gfp_t alloc)return con; } -static int addr_compare (const struct sockaddr_storage *x, -const struct sockaddr_storage *y) +static bool addr_compare (const struct sockaddr_storage *x, +const struct sockaddr_storage *y) {switch (x->ss_family) {case AF_INET : {struct sockaddr_in *sinx = (struct sockaddr_in *)x;struct sockaddr_in *siny = (struct sockaddr_in *)y;if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr) -return 0 ; +return false;if (sinx->sin_port != siny->sin_port) -return 0 ; +return false;break ; }case AF_INET6 : {struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;if (!ipv6_addr_equal (&sinx->sin6_addr, &siny->sin6_addr)) -return 0 ; +return false;if (sinx->sin6_port != siny->sin6_port) -return 0 ; +return false;break ; }default : -return 0 ; +return false; } -return 1 ; +return true; } static int nodeid_to_addr(int nodeid, struct sockaddr_storage *sas_out,
fs/dlm/member.c +9 -9
@@ -18,11 +18,11 @@#include "midcomms.h" #include "lowcomms.h" -int dlm_slots_version (const struct dlm_header *h) +bool dlm_slots_version (const struct dlm_header *h) {if ((le32_to_cpu (h->h_version) &0x0000FFFF ) <DLM_HEADER_SLOTS ) -return 0 ; -return 1 ; +return false; +return true; } void dlm_slot_save(struct dlm_ls *ls, struct dlm_rcom *rc,
@@ -353,20 +353,20 @@ static struct dlm_member *find_memb(struct list_head *head, int nodeid)return NULL ; } -int dlm_is_member (struct dlm_ls *ls,int nodeid) +bool dlm_is_member (struct dlm_ls *ls,int nodeid) {if (find_memb (&ls->ls_nodes, nodeid)) -return 1 ; -return 0 ; +return true; +return false; } -int dlm_is_removed (struct dlm_ls *ls,int nodeid) +bool dlm_is_removed (struct dlm_ls *ls,int nodeid) {WARN_ON_ONCE (!nodeid || nodeid ==-1 );if (find_memb (&ls->ls_nodes_gone, nodeid)) -return 1 ; -return 0 ; +return true; +return false; } static void clear_memb_list(struct list_head *head,
fs/dlm/member.h +3 -3
@@ -16,9 +16,9 @@ int dlm_ls_start(struct dlm_ls *ls);void dlm_clear_members (struct dlm_ls *ls);void dlm_clear_members_gone (struct dlm_ls *ls);int dlm_recover_members (struct dlm_ls *ls,struct dlm_recover *rv,int *neg_out); -int dlm_is_removed (struct dlm_ls *ls,int nodeid); -int dlm_is_member (struct dlm_ls *ls,int nodeid); -int dlm_slots_version (const struct dlm_header *h); +bool dlm_is_removed (struct dlm_ls *ls,int nodeid); +bool dlm_is_member (struct dlm_ls *ls,int nodeid); +bool dlm_slots_version (const struct dlm_header *h);void dlm_slot_save (struct dlm_ls *ls,struct dlm_rcom *rc,struct dlm_member *memb);void dlm_slots_copy_out (struct dlm_ls *ls,struct dlm_rcom *rc);
fs/dlm/requestqueue.c +6 -6
@@ -105,16 +105,16 @@ int dlm_process_requestqueue(struct dlm_ls *ls)return error; } -static int purge_request (struct dlm_ls *ls,struct dlm_message *ms,int nodeid) +static bool purge_request (struct dlm_ls *ls,struct dlm_message *ms,int nodeid) {__le32 type = ms->m_type;/* the ls is being cleaned up and freed by release_lockspace */ if (!atomic_read (&ls->ls_count)) -return 1 ; +return true;if (dlm_is_removed (ls, nodeid)) -return 1 ; +return true;/* directory operations are always purged because the directory is always rebuilt during recovery and the lookups resent */
@@ -122,12 +122,12 @@ static int purge_request(struct dlm_ls *ls, struct dlm_message *ms, int nodeid)if (type ==cpu_to_le32 (DLM_MSG_REMOVE ) || type ==cpu_to_le32 (DLM_MSG_LOOKUP ) || type ==cpu_to_le32 (DLM_MSG_LOOKUP_REPLY )) -return 1 ; +return true;if (!dlm_no_directory (ls)) -return 0 ; +return false; -return 1 ; +return true; }void dlm_purge_requestqueue (struct dlm_ls *ls)
fs/dlm/user.c +8 -8
@@ -156,20 +156,20 @@ static void compat_output(struct dlm_lock_result *res,not related to the lifetime of the lkb struct which is managed entirely by refcount. */ -static int lkb_is_endoflife (int mode,int status) +static bool lkb_is_endoflife (int mode,int status) {switch (status) {case -DLM_EUNLOCK : -return 1 ; +return true;case -DLM_ECANCEL :case -ETIMEDOUT :case -EDEADLK :case -EAGAIN :if (mode ==DLM_LOCK_IV ) -return 1 ; +return true;break ; } -return 0 ; +return false; }/* we could possibly check if the cancel of an orphan has resulted in the lkb
@@ -877,13 +877,13 @@ static __poll_t device_poll(struct file *file, poll_table *wait)return 0 ; } -int dlm_user_daemon_available (void ) +bool dlm_user_daemon_available (void ) {/* dlm_controld hasn't started (or, has started, but not properly populated configfs) */ if (!dlm_our_nodeid ()) -return 0 ; +return false;/* This is to deal with versions of dlm_controld that don't know about the monitor device. We assume that if the
@@ -892,9 +892,9 @@ int dlm_user_daemon_available(void)should open the monitor device before populating configfs. */ if (dlm_monitor_unused) -return 1 ; +return true; -return atomic_read (&dlm_monitor_opened) ?1 :0 ; +return !!atomic_read (&dlm_monitor_opened); }static int ctl_device_open (struct inode *inode,struct file *file)
fs/dlm/user.h +1 -1
@@ -12,6 +12,6 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, uint32_t flags, int mode,int dlm_user_init (void );void dlm_user_exit (void );int dlm_device_deregister (struct dlm_ls *ls); -int dlm_user_daemon_available (void ); +bool dlm_user_daemon_available (void );#endif
Signature · 1 line
2.43.0
Alexander Aring
maintainerPATCH 7/7 dlm: use coniditon expression instead return scalars
In reply to Alexander AringRemoves the condition and the scalar returns, instead we return directly the boolean expression.
--- fs/dlm/lock.c | 47 ++++++++++++------------------------------- fs/dlm/member.c | 12 +++-------- fs/dlm/requestqueue.c | 5 +---- 3 files changed, 17 insertions(+), 47 deletions(-)
fs/dlm/lock.c +13 -34
@@ -260,10 +260,8 @@ static inline bool is_master_copy(struct dlm_lkb *lkb)static inline bool middle_conversion (struct dlm_lkb *lkb) { -if ((lkb->lkb_grmode==DLM_LOCK_PR && lkb->lkb_rqmode==DLM_LOCK_CW ) || - (lkb->lkb_rqmode==DLM_LOCK_PR && lkb->lkb_grmode==DLM_LOCK_CW )) -return true; -return false; +return ((lkb->lkb_grmode ==DLM_LOCK_PR && lkb->lkb_rqmode ==DLM_LOCK_CW ) || + (lkb->lkb_rqmode ==DLM_LOCK_PR && lkb->lkb_grmode ==DLM_LOCK_CW )); }static inline bool down_conversion (struct dlm_lkb *lkb)
@@ -2140,10 +2138,7 @@ static inline bool first_in_list(struct dlm_lkb *lkb, struct list_head *head) { struct dlm_lkb *first = list_entry(head->next, struct dlm_lkb, lkb_statequeue); -if (lkb->lkb_id == first->lkb_id) -return true; - -return false; +return (lkb->lkb_id == first->lkb_id); }/* Check if the given lkb conflicts with another lkb on the queue. */
@@ -2380,11 +2375,8 @@ static bool _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, bool now,* of the most restrictive lock granted on the resource. */ -if (!now && !conv &&list_empty (&r->res_convertqueue) && -first_in_list (lkb, &r->res_waitqueue)) -return true; - -return false; +return (!now && !conv &&list_empty (&r->res_convertqueue) && +first_in_list (lkb, &r->res_waitqueue)); } static bool can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, bool now,
@@ -2544,16 +2536,11 @@ static int grant_pending_wait(struct dlm_rsb *r, int high, bool *cw,static bool lock_requires_bast (struct dlm_lkb *gr,int high,bool cw) { -if (gr->lkb_grmode ==DLM_LOCK_PR && cw) { -if (gr->lkb_highbast <DLM_LOCK_EX ) -return true; -return false; - } +if (gr->lkb_grmode ==DLM_LOCK_PR && cw) +return (gr->lkb_highbast <DLM_LOCK_EX ); -if (gr->lkb_highbast < high && - !__dlm_compat_matrix[gr->lkb_grmode+1 ][high+1 ]) -return true; -return false; +return (gr->lkb_highbast < high && + !__dlm_compat_matrix[gr->lkb_grmode+1 ][high+1 ]); }static void grant_pending_locks (struct dlm_rsb *r,unsigned int *count)
@@ -2595,15 +2582,10 @@ static void grant_pending_locks(struct dlm_rsb *r, unsigned int *count)static int modes_require_bast (struct dlm_lkb *gr,struct dlm_lkb *rq) {if ((gr->lkb_grmode ==DLM_LOCK_PR && rq->lkb_rqmode ==DLM_LOCK_CW ) || - (gr->lkb_grmode ==DLM_LOCK_CW && rq->lkb_rqmode ==DLM_LOCK_PR )) { -if (gr->lkb_highbast <DLM_LOCK_EX ) -return true; -return false; - } + (gr->lkb_grmode ==DLM_LOCK_CW && rq->lkb_rqmode ==DLM_LOCK_PR )) +return (gr->lkb_highbast <DLM_LOCK_EX ); -if (gr->lkb_highbast < rq->lkb_rqmode && !modes_compat (gr, rq)) -return true; -return false; +return (gr->lkb_highbast < rq->lkb_rqmode && !modes_compat (gr, rq)); } static void send_bast_queue(struct dlm_rsb *r, struct list_head *head,
@@ -5030,10 +5012,7 @@ static bool waiter_needs_recovery(struct dlm_ls *ls, struct dlm_lkb *lkb,if (dlm_no_directory (ls))return true; -if (dlm_is_removed (ls, lkb->lkb_wait_nodeid)) -return true; - -return false; +return dlm_is_removed (ls, lkb->lkb_wait_nodeid); }/* Recovery for locks that are waiting for replies from nodes that are now
fs/dlm/member.c +3 -9
@@ -20,9 +20,7 @@bool dlm_slots_version (const struct dlm_header *h) { -if ((le32_to_cpu (h->h_version) &0x0000FFFF ) <DLM_HEADER_SLOTS ) -return false; -return true; +return !((le32_to_cpu (h->h_version) &0x0000FFFF ) <DLM_HEADER_SLOTS ); } void dlm_slot_save(struct dlm_ls *ls, struct dlm_rcom *rc,
@@ -355,18 +353,14 @@ static struct dlm_member *find_memb(struct list_head *head, int nodeid)bool dlm_is_member (struct dlm_ls *ls,int nodeid) { -if (find_memb (&ls->ls_nodes, nodeid)) -return true; -return false; +return !!find_memb (&ls->ls_nodes, nodeid); }bool dlm_is_removed (struct dlm_ls *ls,int nodeid) {WARN_ON_ONCE (!nodeid || nodeid ==-1 ); -if (find_memb (&ls->ls_nodes_gone, nodeid)) -return true; -return false; +return !!find_memb (&ls->ls_nodes_gone, nodeid); } static void clear_memb_list(struct list_head *head,
fs/dlm/requestqueue.c +1 -4
@@ -124,10 +124,7 @@ static bool purge_request(struct dlm_ls *ls, struct dlm_message *ms, int nodeid) type == cpu_to_le32(DLM_MSG_LOOKUP_REPLY))return true; -if (!dlm_no_directory (ls)) -return false; - -return true; +return dlm_no_directory (ls); }void dlm_purge_requestqueue (struct dlm_ls *ls)
Signature · 1 line
2.43.0