[PATCH] v9fs: signal handling fixes
authorLatchesar Ionkov <lucho@ionkov.net>
Mon, 15 May 2006 16:44:21 +0000 (09:44 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Mon, 15 May 2006 18:20:56 +0000 (11:20 -0700)
Multiple races can happen when v9fs is interrupted by a signal and Tflush
message is sent to the server.  After v9fs sends Tflush it doesn't wait
until it receives Rflush, and possibly the response of the original
message.  This behavior may confuse v9fs what fids are allocated by the
file server.

This patch fixes the races and the fid allocation.

Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Cc: Eric Van Hensbergen <ericvh@hera.kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/9p/mux.c
fs/9p/mux.h
fs/9p/vfs_file.c
fs/9p/vfs_inode.c

index 3e5b124..f4407eb 100644 (file)
@@ -50,15 +50,23 @@ enum {
        Wpending = 8,           /* can write */
 };
 
+enum {
+       None,
+       Flushing,
+       Flushed,
+};
+
 struct v9fs_mux_poll_task;
 
 struct v9fs_req {
+       spinlock_t lock;
        int tag;
        struct v9fs_fcall *tcall;
        struct v9fs_fcall *rcall;
        int err;
        v9fs_mux_req_callback cb;
        void *cba;
+       int flush;
        struct list_head req_list;
 };
 
@@ -96,8 +104,8 @@ struct v9fs_mux_poll_task {
 
 struct v9fs_mux_rpc {
        struct v9fs_mux_data *m;
-       struct v9fs_req *req;
        int err;
+       struct v9fs_fcall *tcall;
        struct v9fs_fcall *rcall;
        wait_queue_head_t wqueue;
 };
@@ -524,10 +532,9 @@ again:
 
 static void process_request(struct v9fs_mux_data *m, struct v9fs_req *req)
 {
-       int ecode, tag;
+       int ecode;
        struct v9fs_str *ename;
 
-       tag = req->tag;
        if (!req->err && req->rcall->id == RERROR) {
                ecode = req->rcall->params.rerror.errno;
                ename = &req->rcall->params.rerror.error;
@@ -553,23 +560,6 @@ static void process_request(struct v9fs_mux_data *m, struct v9fs_req *req)
                if (!req->err)
                        req->err = -EIO;
        }
-
-       if (req->err == ERREQFLUSH)
-               return;
-
-       if (req->cb) {
-               dprintk(DEBUG_MUX, "calling callback tcall %p rcall %p\n",
-                       req->tcall, req->rcall);
-
-               (*req->cb) (req->cba, req->tcall, req->rcall, req->err);
-               req->cb = NULL;
-       } else
-               kfree(req->rcall);
-
-       v9fs_mux_put_tag(m, tag);
-
-       wake_up(&m->equeue);
-       kfree(req);
 }
 
 /**
@@ -669,17 +659,26 @@ static void v9fs_read_work(void *a)
                list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
                        if (rreq->tag == rcall->tag) {
                                req = rreq;
-                               req->rcall = rcall;
-                               list_del(&req->req_list);
-                               spin_unlock(&m->lock);
-                               process_request(m, req);
+                               if (req->flush != Flushing)
+                                       list_del(&req->req_list);
                                break;
                        }
-
                }
+               spin_unlock(&m->lock);
 
-               if (!req) {
-                       spin_unlock(&m->lock);
+               if (req) {
+                       req->rcall = rcall;
+                       process_request(m, req);
+
+                       if (req->flush != Flushing) {
+                               if (req->cb)
+                                       (*req->cb) (req, req->cba);
+                               else
+                                       kfree(req->rcall);
+
+                               wake_up(&m->equeue);
+                       }
+               } else {
                        if (err >= 0 && rcall->id != RFLUSH)
                                dprintk(DEBUG_ERROR,
                                        "unexpected response mux %p id %d tag %d\n",
@@ -746,7 +745,6 @@ static struct v9fs_req *v9fs_send_request(struct v9fs_mux_data *m,
                return ERR_PTR(-ENOMEM);
 
        v9fs_set_tag(tc, n);
-
        if ((v9fs_debug_level&DEBUG_FCALL) == DEBUG_FCALL) {
                char buf[150];
 
@@ -754,12 +752,14 @@ static struct v9fs_req *v9fs_send_request(struct v9fs_mux_data *m,
                printk(KERN_NOTICE "<<< %p %s\n", m, buf);
        }
 
+       spin_lock_init(&req->lock);
        req->tag = n;
        req->tcall = tc;
        req->rcall = NULL;
        req->err = 0;
        req->cb = cb;
        req->cba = cba;
+       req->flush = None;
 
        spin_lock(&m->lock);
        list_add_tail(&req->req_list, &m->unsent_req_list);
@@ -776,72 +776,108 @@ static struct v9fs_req *v9fs_send_request(struct v9fs_mux_data *m,
        return req;
 }
 
-static void v9fs_mux_flush_cb(void *a, struct v9fs_fcall *tc,
-                             struct v9fs_fcall *rc, int err)
+static void v9fs_mux_free_request(struct v9fs_mux_data *m, struct v9fs_req *req)
+{
+       v9fs_mux_put_tag(m, req->tag);
+       kfree(req);
+}
+
+static void v9fs_mux_flush_cb(struct v9fs_req *freq, void *a)
 {
        v9fs_mux_req_callback cb;
        int tag;
        struct v9fs_mux_data *m;
-       struct v9fs_req *req, *rptr;
+       struct v9fs_req *req, *rreq, *rptr;
 
        m = a;
-       dprintk(DEBUG_MUX, "mux %p tc %p rc %p err %d oldtag %d\n", m, tc,
-               rc, err, tc->params.tflush.oldtag);
+       dprintk(DEBUG_MUX, "mux %p tc %p rc %p err %d oldtag %d\n", m,
+               freq->tcall, freq->rcall, freq->err,
+               freq->tcall->params.tflush.oldtag);
 
        spin_lock(&m->lock);
        cb = NULL;
-       tag = tc->params.tflush.oldtag;
-       list_for_each_entry_safe(req, rptr, &m->req_list, req_list) {
-               if (req->tag == tag) {
+       tag = freq->tcall->params.tflush.oldtag;
+       req = NULL;
+       list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
+               if (rreq->tag == tag) {
+                       req = rreq;
                        list_del(&req->req_list);
-                       if (req->cb) {
-                               cb = req->cb;
-                               req->cb = NULL;
-                               spin_unlock(&m->lock);
-                               (*cb) (req->cba, req->tcall, req->rcall,
-                                      req->err);
-                       }
-                       kfree(req);
-                       wake_up(&m->equeue);
                        break;
                }
        }
+       spin_unlock(&m->lock);
 
-       if (!cb)
-               spin_unlock(&m->lock);
+       if (req) {
+               spin_lock(&req->lock);
+               req->flush = Flushed;
+               spin_unlock(&req->lock);
+
+               if (req->cb)
+                       (*req->cb) (req, req->cba);
+               else
+                       kfree(req->rcall);
+
+               wake_up(&m->equeue);
+       }
 
-       v9fs_mux_put_tag(m, tag);
-       kfree(tc);
-       kfree(rc);
+       kfree(freq->tcall);
+       kfree(freq->rcall);
+       v9fs_mux_free_request(m, freq);
 }
 
-static void
+static int
 v9fs_mux_flush_request(struct v9fs_mux_data *m, struct v9fs_req *req)
 {
        struct v9fs_fcall *fc;
+       struct v9fs_req *rreq, *rptr;
 
        dprintk(DEBUG_MUX, "mux %p req %p tag %d\n", m, req, req->tag);
 
+       /* if a response was received for a request, do nothing */
+       spin_lock(&req->lock);
+       if (req->rcall || req->err) {
+               spin_unlock(&req->lock);
+               dprintk(DEBUG_MUX, "mux %p req %p response already received\n", m, req);
+               return 0;
+       }
+
+       req->flush = Flushing;
+       spin_unlock(&req->lock);
+
+       spin_lock(&m->lock);
+       /* if the request is not sent yet, just remove it from the list */
+       list_for_each_entry_safe(rreq, rptr, &m->unsent_req_list, req_list) {
+               if (rreq->tag == req->tag) {
+                       dprintk(DEBUG_MUX, "mux %p req %p request is not sent yet\n", m, req);
+                       list_del(&rreq->req_list);
+                       req->flush = Flushed;
+                       spin_unlock(&m->lock);
+                       if (req->cb)
+                               (*req->cb) (req, req->cba);
+                       return 0;
+               }
+       }
+       spin_unlock(&m->lock);
+
+       clear_thread_flag(TIF_SIGPENDING);
        fc = v9fs_create_tflush(req->tag);
        v9fs_send_request(m, fc, v9fs_mux_flush_cb, m);
+       return 1;
 }
 
 static void
-v9fs_mux_rpc_cb(void *a, struct v9fs_fcall *tc, struct v9fs_fcall *rc, int err)
+v9fs_mux_rpc_cb(struct v9fs_req *req, void *a)
 {
        struct v9fs_mux_rpc *r;
 
-       if (err == ERREQFLUSH) {
-               kfree(rc);
-               dprintk(DEBUG_MUX, "err req flush\n");
-               return;
-       }
-
+       dprintk(DEBUG_MUX, "req %p r %p\n", req, a);
        r = a;
-       dprintk(DEBUG_MUX, "mux %p req %p tc %p rc %p err %d\n", r->m, r->req,
-               tc, rc, err);
-       r->rcall = rc;
-       r->err = err;
+       r->rcall = req->rcall;
+       r->err = req->err;
+
+       if (req->flush!=None && !req->err)
+               r->err = -ERESTARTSYS;
+
        wake_up(&r->wqueue);
 }
 
@@ -856,12 +892,13 @@ int
 v9fs_mux_rpc(struct v9fs_mux_data *m, struct v9fs_fcall *tc,
             struct v9fs_fcall **rc)
 {
-       int err;
+       int err, sigpending;
        unsigned long flags;
        struct v9fs_req *req;
        struct v9fs_mux_rpc r;
 
        r.err = 0;
+       r.tcall = tc;
        r.rcall = NULL;
        r.m = m;
        init_waitqueue_head(&r.wqueue);
@@ -869,48 +906,50 @@ v9fs_mux_rpc(struct v9fs_mux_data *m, struct v9fs_fcall *tc,
        if (rc)
                *rc = NULL;
 
+       sigpending = 0;
+       if (signal_pending(current)) {
+               sigpending = 1;
+               clear_thread_flag(TIF_SIGPENDING);
+       }
+
        req = v9fs_send_request(m, tc, v9fs_mux_rpc_cb, &r);
        if (IS_ERR(req)) {
                err = PTR_ERR(req);
                dprintk(DEBUG_MUX, "error %d\n", err);
-               return PTR_ERR(req);
+               return err;
        }
 
-       r.req = req;
-       dprintk(DEBUG_MUX, "mux %p tc %p tag %d rpc %p req %p\n", m, tc,
-               req->tag, &r, req);
        err = wait_event_interruptible(r.wqueue, r.rcall != NULL || r.err < 0);
        if (r.err < 0)
                err = r.err;
 
        if (err == -ERESTARTSYS && m->trans->status == Connected && m->err == 0) {
-               spin_lock(&m->lock);
-               req->tcall = NULL;
-               req->err = ERREQFLUSH;
-               spin_unlock(&m->lock);
+               if (v9fs_mux_flush_request(m, req)) {
+                       /* wait until we get response of the flush message */
+                       do {
+                               clear_thread_flag(TIF_SIGPENDING);
+                               err = wait_event_interruptible(r.wqueue,
+                                       r.rcall || r.err);
+                       } while (!r.rcall && !r.err && err==-ERESTARTSYS &&
+                               m->trans->status==Connected && !m->err);
+               }
+               sigpending = 1;
+       }
 
-               clear_thread_flag(TIF_SIGPENDING);
-               v9fs_mux_flush_request(m, req);
+       if (sigpending) {
                spin_lock_irqsave(&current->sighand->siglock, flags);
                recalc_sigpending();
                spin_unlock_irqrestore(&current->sighand->siglock, flags);
        }
 
-       if (!err) {
-               if (r.rcall)
-                       dprintk(DEBUG_MUX, "got response id %d tag %d\n",
-                               r.rcall->id, r.rcall->tag);
-
-               if (rc)
-                       *rc = r.rcall;
-               else
-                       kfree(r.rcall);
-       } else {
+       if (rc)
+               *rc = r.rcall;
+       else
                kfree(r.rcall);
-               dprintk(DEBUG_MUX, "got error %d\n", err);
-               if (err > 0)
-                       err = -EIO;
-       }
+
+       v9fs_mux_free_request(m, req);
+       if (err > 0)
+               err = -EIO;
 
        return err;
 }
@@ -951,12 +990,15 @@ void v9fs_mux_cancel(struct v9fs_mux_data *m, int err)
        struct v9fs_req *req, *rtmp;
        LIST_HEAD(cancel_list);
 
-       dprintk(DEBUG_MUX, "mux %p err %d\n", m, err);
+       dprintk(DEBUG_ERROR, "mux %p err %d\n", m, err);
        m->err = err;
        spin_lock(&m->lock);
        list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
                list_move(&req->req_list, &cancel_list);
        }
+       list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
+               list_move(&req->req_list, &cancel_list);
+       }
        spin_unlock(&m->lock);
 
        list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
@@ -965,11 +1007,9 @@ void v9fs_mux_cancel(struct v9fs_mux_data *m, int err)
                        req->err = err;
 
                if (req->cb)
-                       (*req->cb) (req->cba, req->tcall, req->rcall, req->err);
+                       (*req->cb) (req, req->cba);
                else
                        kfree(req->rcall);
-
-               kfree(req);
        }
 
        wake_up(&m->equeue);
index e90bfd3..fb10c50 100644 (file)
@@ -24,6 +24,7 @@
  */
 
 struct v9fs_mux_data;
+struct v9fs_req;
 
 /**
  * v9fs_mux_req_callback - callback function that is called when the
@@ -36,8 +37,7 @@ struct v9fs_mux_data;
  * @rc - response call
  * @err - error code (non-zero if error occured)
  */
-typedef void (*v9fs_mux_req_callback)(void *a, struct v9fs_fcall *tc,
-       struct v9fs_fcall *rc, int err);
+typedef void (*v9fs_mux_req_callback)(struct v9fs_req *req, void *a);
 
 int v9fs_mux_global_init(void);
 void v9fs_mux_global_exit(void);
index 083dcfc..1a8e460 100644 (file)
@@ -72,11 +72,17 @@ int v9fs_file_open(struct inode *inode, struct file *file)
                return -ENOSPC;
        }
 
-       err = v9fs_t_walk(v9ses, vfid->fid, fid, NULL, NULL);
+       err = v9fs_t_walk(v9ses, vfid->fid, fid, NULL, &fcall);
        if (err < 0) {
                dprintk(DEBUG_ERROR, "rewalk didn't work\n");
-               goto put_fid;
+               if (fcall && fcall->id == RWALK)
+                       goto clunk_fid;
+               else {
+                       v9fs_put_idpool(fid, &v9ses->fidpool);
+                       goto free_fcall;
+               }
        }
+       kfree(fcall);
 
        /* TODO: do special things for O_EXCL, O_NOFOLLOW, O_SYNC */
        /* translate open mode appropriately */
@@ -109,8 +115,7 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 clunk_fid:
        v9fs_t_clunk(v9ses, fid);
 
-put_fid:
-       v9fs_put_idpool(fid, &v9ses->fidpool);
+free_fcall:
        kfree(fcall);
 
        return err;
index 133db36..2cb87ba 100644 (file)
@@ -270,7 +270,10 @@ v9fs_create(struct v9fs_session_info *v9ses, u32 pfid, char *name, u32 perm,
        err = v9fs_t_walk(v9ses, pfid, fid, NULL, &fcall);
        if (err < 0) {
                PRINT_FCALL_ERROR("clone error", fcall);
-               goto put_fid;
+               if (fcall && fcall->id == RWALK)
+                       goto clunk_fid;
+               else
+                       goto put_fid;
        }
        kfree(fcall);
 
@@ -322,6 +325,9 @@ v9fs_clone_walk(struct v9fs_session_info *v9ses, u32 fid, struct dentry *dentry)
                &fcall);
 
        if (err < 0) {
+               if (fcall && fcall->id == RWALK)
+                       goto clunk_fid;
+
                PRINT_FCALL_ERROR("walk error", fcall);
                v9fs_put_idpool(nfid, &v9ses->fidpool);
                goto error;
@@ -640,19 +646,26 @@ static struct dentry *v9fs_vfs_lookup(struct inode *dir, struct dentry *dentry,
        }
 
        result = v9fs_t_walk(v9ses, dirfidnum, newfid,
-               (char *)dentry->d_name.name, NULL);
+               (char *)dentry->d_name.name, &fcall);
+
        if (result < 0) {
-               v9fs_put_idpool(newfid, &v9ses->fidpool);
+               if (fcall && fcall->id == RWALK)
+                       v9fs_t_clunk(v9ses, newfid);
+               else
+                       v9fs_put_idpool(newfid, &v9ses->fidpool);
+
                if (result == -ENOENT) {
                        d_add(dentry, NULL);
                        dprintk(DEBUG_VFS,
                                "Return negative dentry %p count %d\n",
                                dentry, atomic_read(&dentry->d_count));
+                       kfree(fcall);
                        return NULL;
                }
                dprintk(DEBUG_ERROR, "walk error:%d\n", result);
                goto FreeFcall;
        }
+       kfree(fcall);
 
        result = v9fs_t_stat(v9ses, newfid, &fcall);
        if (result < 0) {