hw/9pfs: Add file descriptor reclaim support

[M. Mohan Kumar <mohan@in.ibm.com> removed some unused variables]

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

hw/9pfs/codir.c

@@ -100,12 +100,10 @@ int v9fs_co_opendir(V9fsState *s, V9fsFidState *fidp)
     return err;
 }
 
-int v9fs_co_closedir(V9fsState *s, V9fsFidState *fidp)
+int v9fs_co_closedir(V9fsState *s, DIR *dir)
 {
     int err;
-    DIR *dir;
 
-    dir = fidp->fs.dir;
     v9fs_co_run_in_worker(
         {
             err = s->ops->closedir(&s->ctx, dir);

hw/9pfs/cofile.c

@@ -58,6 +58,12 @@ int v9fs_co_open(V9fsState *s, V9fsFidState *fidp, int flags)
                 err = 0;
             }
         });
+    if (!err) {
+        total_open_fd++;
+        if (total_open_fd > open_fd_hw) {
+            v9fs_reclaim_fd(s);
+        }
+    }
     return err;
 }
 
@@ -79,15 +85,19 @@ int v9fs_co_open2(V9fsState *s, V9fsFidState *fidp, char *fullname, gid_t gid,
                 err = -errno;
             }
         });
+    if (!err) {
+        total_open_fd++;
+        if (total_open_fd > open_fd_hw) {
+            v9fs_reclaim_fd(s);
+        }
+    }
     return err;
 }
 
-int v9fs_co_close(V9fsState *s, V9fsFidState *fidp)
+int v9fs_co_close(V9fsState *s, int fd)
 {
-    int fd;
     int err;
 
-    fd = fidp->fs.fd;
     v9fs_co_run_in_worker(
         {
             err = s->ops->close(&s->ctx, fd);
@@ -95,6 +105,9 @@ int v9fs_co_close(V9fsState *s, V9fsFidState *fidp)
                 err = -errno;
             }
         });
+    if (!err) {
+        total_open_fd--;
+    }
     return err;
 }
 

hw/9pfs/virtio-9p-coth.h

@@ -83,8 +83,8 @@ extern int v9fs_co_open2(V9fsState *, V9fsFidState *, char *, gid_t, int, int);
 extern int v9fs_co_lsetxattr(V9fsState *, V9fsString *, V9fsString *,
                              void *, size_t, int);
 extern int v9fs_co_lremovexattr(V9fsState *, V9fsString *, V9fsString *);
-extern int v9fs_co_closedir(V9fsState *, V9fsFidState *);
+extern int v9fs_co_closedir(V9fsState *, DIR *);
-extern int v9fs_co_close(V9fsState *, V9fsFidState *);
+extern int v9fs_co_close(V9fsState *, int);
 extern int v9fs_co_fsync(V9fsState *, V9fsFidState *, int);
 extern int v9fs_co_symlink(V9fsState *, V9fsFidState *, const char *,
                            const char *, gid_t);

hw/9pfs/virtio-9p-device.c

@@ -171,6 +171,7 @@ static PCIDeviceInfo virtio_9p_info = {
 static void virtio_9p_register_devices(void)
 {
     pci_qdev_register(&virtio_9p_info);
+    virtio_9p_set_fd_limit();
 }
 
 device_init(virtio_9p_register_devices)

hw/9pfs/virtio-9p.c

@@ -22,6 +22,9 @@
 #include "virtio-9p-coth.h"
 
 int debug_9p_pdu;
+int open_fd_hw;
+int total_open_fd;
+static int open_fd_rc;
 
 enum {
     Oread   = 0x00,
@@ -234,12 +237,41 @@ static size_t v9fs_string_size(V9fsString *str)
 
 static V9fsFidState *get_fid(V9fsState *s, int32_t fid)
 {
+    int err;
     V9fsFidState *f;
 
     for (f = s->fid_list; f; f = f->next) {
         BUG_ON(f->clunked);
         if (f->fid == fid) {
+            /*
+             * Update the fid ref upfront so that
+             * we don't get reclaimed when we yield
+             * in open later.
+             */
             f->ref++;
+
+            /*
+             * check whether we need to reopen the
+             * file. We might have closed the fd
+             * while trying to free up some file
+             * descriptors.
+             */
+            if (f->fid_type == P9_FID_FILE) {
+                if (f->fs.fd == -1) {
+                    do {
+                        err = v9fs_co_open(s, f, f->open_flags);
+                    } while (err == -EINTR);
+                    if (err < 0) {
+                        f->ref--;
+                        return NULL;
+                    }
+                }
+            }
+            /*
+             * Mark the fid as referenced so that the LRU
+             * reclaim won't close the file descriptor
+             */
+            f->flags |= FID_REFERENCED;
             return f;
         }
     }
@@ -261,6 +293,11 @@ static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
     f->fid = fid;
     f->fid_type = P9_FID_NONE;
     f->ref = 1;
+    /*
+     * Mark the fid as referenced so that the LRU
+     * reclaim won't close the file descriptor
+     */
+    f->flags |= FID_REFERENCED;
     f->next = s->fid_list;
     s->fid_list = f;
 
@@ -306,9 +343,12 @@ static int free_fid(V9fsState *s, V9fsFidState *fidp)
     int retval = 0;
 
     if (fidp->fid_type == P9_FID_FILE) {
-        retval = v9fs_co_close(s, fidp);
+        /* If we reclaimed the fd no need to close */
+        if (fidp->fs.fd != -1) {
+            retval = v9fs_co_close(s, fidp->fs.fd);
+        }
     } else if (fidp->fid_type == P9_FID_DIR) {
-        retval = v9fs_co_closedir(s, fidp);
+        retval = v9fs_co_closedir(s, fidp->fs.dir);
     } else if (fidp->fid_type == P9_FID_XATTR) {
         retval = v9fs_xattr_fid_clunk(s, fidp);
     }
@@ -321,6 +361,9 @@ static void put_fid(V9fsState *s, V9fsFidState *fidp)
 {
     BUG_ON(!fidp->ref);
     fidp->ref--;
+    /*
+     * Don't free the fid if it is in reclaim list
+     */
     if (!fidp->ref && fidp->clunked) {
         free_fid(s, fidp);
     }
@@ -345,6 +388,101 @@ static int clunk_fid(V9fsState *s, int32_t fid)
     return 0;
 }
 
+void v9fs_reclaim_fd(V9fsState *s)
+{
+    int reclaim_count = 0;
+    V9fsFidState *f, *reclaim_list = NULL;
+
+    for (f = s->fid_list; f; f = f->next) {
+        /*
+         * Unlink fids cannot be reclaimed. Check
+         * for them and skip them. Also skip fids
+         * currently being operated on.
+         */
+        if (f->ref || f->flags & FID_NON_RECLAIMABLE) {
+            continue;
+        }
+        /*
+         * if it is a recently referenced fid
+         * we leave the fid untouched and clear the
+         * reference bit. We come back to it later
+         * in the next iteration. (a simple LRU without
+         * moving list elements around)
+         */
+        if (f->flags & FID_REFERENCED) {
+            f->flags &= ~FID_REFERENCED;
+            continue;
+        }
+        /*
+         * Add fids to reclaim list.
+         */
+        if (f->fid_type == P9_FID_FILE) {
+            if (f->fs.fd != -1) {
+                /*
+                 * Up the reference count so that
+                 * a clunk request won't free this fid
+                 */
+                f->ref++;
+                f->rclm_lst = reclaim_list;
+                reclaim_list = f;
+                f->fs_reclaim.fd = f->fs.fd;
+                f->fs.fd = -1;
+                reclaim_count++;
+            }
+        }
+        if (reclaim_count >= open_fd_rc) {
+            break;
+        }
+    }
+    /*
+     * Now close the fid in reclaim list. Free them if they
+     * are already clunked.
+     */
+    while (reclaim_list) {
+        f = reclaim_list;
+        reclaim_list = f->rclm_lst;
+        if (f->fid_type == P9_FID_FILE) {
+            v9fs_co_close(s, f->fs_reclaim.fd);
+        }
+        f->rclm_lst = NULL;
+        /*
+         * Now drop the fid reference, free it
+         * if clunked.
+         */
+        put_fid(s, f);
+    }
+}
+
+static int v9fs_mark_fids_unreclaim(V9fsState *s, V9fsString *str)
+{
+    int err;
+    V9fsFidState *fidp, head_fid;
+
+    head_fid.next = s->fid_list;
+    for (fidp = s->fid_list; fidp; fidp = fidp->next) {
+        if (!strcmp(fidp->path.data, str->data)) {
+            /* Mark the fid non reclaimable. */
+            fidp->flags |= FID_NON_RECLAIMABLE;
+            /* reopen the file if already closed */
+            if (fidp->fs.fd == -1) {
+                do {
+                    err = v9fs_co_open(s, fidp, fidp->open_flags);
+                } while (err == -EINTR);
+                if (err < 0) {
+                    return -1;
+                }
+                /*
+                 * Go back to head of fid list because
+                 * the list could have got updated when
+                 * switched to the worker thread
+                 */
+                fidp = &head_fid;
+            }
+        }
+    }
+    return 0;
+}
+
 #define P9_QID_TYPE_DIR         0x80
 #define P9_QID_TYPE_SYMLINK     0x02
 
@@ -1389,6 +1527,14 @@ static void v9fs_open(void *opaque)
             goto out;
         }
         fidp->fid_type = P9_FID_FILE;
+        fidp->open_flags = flags;
+        if (flags & O_EXCL) {
+            /*
+             * We let the host file system do O_EXCL check
+             * We should not reclaim such fd
+             */
+            fidp->flags |= FID_NON_RECLAIMABLE;
+        }
         iounit = get_iounit(s, &fidp->path);
         offset += pdu_marshal(pdu, offset, "Qd", &qid, iounit);
         err = offset;
@@ -1432,6 +1578,14 @@ static void v9fs_lcreate(void *opaque)
         goto out;
     }
     fidp->fid_type = P9_FID_FILE;
+    fidp->open_flags = flags;
+    if (flags & O_EXCL) {
+        /*
+         * We let the host file system do O_EXCL check
+         * We should not reclaim such fd
+         */
+        fidp->flags |= FID_NON_RECLAIMABLE;
+    }
     iounit =  get_iounit(pdu->s, &fullname);
 
     err = v9fs_co_lstat(pdu->s, &fullname, &stbuf);
@@ -1993,6 +2147,14 @@ static void v9fs_create(void *opaque)
             goto out;
         }
         fidp->fid_type = P9_FID_FILE;
+        fidp->open_flags = omode_to_uflags(mode);
+        if (fidp->open_flags & O_EXCL) {
+            /*
+             * We let the host file system do O_EXCL check
+             * We should not reclaim such fd
+             */
+            fidp->flags |= FID_NON_RECLAIMABLE;
+        }
     }
     err = v9fs_co_lstat(pdu->s, &fullname, &stbuf);
     if (err < 0) {
@@ -2124,11 +2286,19 @@ static void v9fs_remove(void *opaque)
         err = -EINVAL;
         goto out_nofid;
     }
+    /*
+     * IF the file is unlinked, we cannot reopen
+     * the file later. So don't reclaim fd
+     */
+    err = v9fs_mark_fids_unreclaim(pdu->s, &fidp->path);
+    if (err < 0) {
+        goto out_err;
+    }
     err = v9fs_co_remove(pdu->s, &fidp->path);
     if (!err) {
         err = offset;
     }
-
+out_err:
     /* For TREMOVE we need to clunk the fid even on failed remove */
     clunk_fid(pdu->s, fidp->fid);
     put_fid(pdu->s, fidp);
@@ -2823,3 +2993,14 @@ void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq)
     }
     free_pdu(s, pdu);
 }
+
+void virtio_9p_set_fd_limit(void)
+{
+    struct rlimit rlim;
+    if (getrlimit(RLIMIT_NOFILE, &rlim) < 0) {
+        fprintf(stderr, "Failed to get the resource limit\n");
+        exit(1);
+    }
+    open_fd_hw = rlim.rlim_cur - MIN(400, rlim.rlim_cur/3);
+    open_fd_rc = rlim.rlim_cur/2;
+}

hw/9pfs/virtio-9p.h

@@ -5,6 +5,7 @@
 #include <dirent.h>
 #include <sys/time.h>
 #include <utime.h>
+#include <sys/resource.h>
 #include "hw/virtio.h"
 #include "fsdev/file-op-9p.h"
 
@@ -101,6 +102,9 @@ enum p9_proto_version {
 #define P9_NOTAG    (u16)(~0)
 #define P9_NOFID    (u32)(~0)
 #define P9_MAXWELEM 16
+
+#define FID_REFERENCED          0x1
+#define FID_NON_RECLAIMABLE     0x2
 static inline const char *rpath(FsContext *ctx, const char *path, char *buffer)
 {
     snprintf(buffer, PATH_MAX, "%s/%s", ctx->fs_root, path);
@@ -198,14 +202,21 @@ struct V9fsFidState
     int32_t fid;
     V9fsString path;
     union {
-	int fd;
+        int fd;
-	DIR *dir;
+        DIR *dir;
-	V9fsXattr xattr;
+        V9fsXattr xattr;
     } fs;
+    union {
+        int fd;
+        DIR *dir;
+    } fs_reclaim;
+    int flags;
+    int open_flags;
     uid_t uid;
     int ref;
     int clunked;
     V9fsFidState *next;
+    V9fsFidState *rclm_lst;
 };
 
 typedef struct V9fsState
@@ -354,6 +365,9 @@ typedef struct V9fsGetlock
     V9fsString client_id;
 } V9fsGetlock;
 
+extern int open_fd_hw;
+extern int total_open_fd;
+
 size_t pdu_packunpack(void *addr, struct iovec *sg, int sg_count,
                       size_t offset, size_t size, int pack);
 
@@ -364,4 +378,6 @@ static inline size_t do_pdu_unpack(void *dst, struct iovec *sg, int sg_count,
 }
 
 extern void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq);
+extern void virtio_9p_set_fd_limit(void);
+extern void v9fs_reclaim_fd(V9fsState *s);
 #endif