From: Mike Marshall Date: Fri, 17 Jul 2015 14:38:13 +0000 (-0400) Subject: Orangefs: kernel client part 3 X-Git-Url: https://git.karo-electronics.de/?a=commitdiff_plain;h=274dcf55bd4ab12af1cc1d3b77416285bef8ebf4;p=linux-beck.git Orangefs: kernel client part 3 Signed-off-by: Mike Marshall --- diff --git a/fs/orangefs/namei.c b/fs/orangefs/namei.c new file mode 100644 index 000000000000..747fe6a690af --- /dev/null +++ b/fs/orangefs/namei.c @@ -0,0 +1,473 @@ +/* + * (C) 2001 Clemson University and The University of Chicago + * + * See COPYING in top-level directory. + */ + +/* + * Linux VFS namei operations. + */ + +#include "protocol.h" +#include "pvfs2-kernel.h" + +/* + * Get a newly allocated inode to go with a negative dentry. + */ +static int pvfs2_create(struct inode *dir, + struct dentry *dentry, + umode_t mode, + bool exclusive) +{ + struct pvfs2_inode_s *parent = PVFS2_I(dir); + struct pvfs2_kernel_op_s *new_op; + struct inode *inode; + int ret; + + gossip_debug(GOSSIP_NAME_DEBUG, "%s: called\n", __func__); + + new_op = op_alloc(PVFS2_VFS_OP_CREATE); + if (!new_op) + return -ENOMEM; + + new_op->upcall.req.create.parent_refn = parent->refn; + + fill_default_sys_attrs(new_op->upcall.req.create.attributes, + PVFS_TYPE_METAFILE, mode); + + strncpy(new_op->upcall.req.create.d_name, + dentry->d_name.name, PVFS2_NAME_LEN); + + ret = service_operation(new_op, __func__, get_interruptible_flag(dir)); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Create Got PVFS2 handle %pU on fsid %d (ret=%d)\n", + &new_op->downcall.resp.create.refn.khandle, + new_op->downcall.resp.create.refn.fs_id, ret); + + if (ret < 0) { + gossip_debug(GOSSIP_NAME_DEBUG, + "%s: failed with error code %d\n", + __func__, ret); + goto out; + } + + inode = pvfs2_new_inode(dir->i_sb, dir, S_IFREG | mode, 0, + &new_op->downcall.resp.create.refn); + if (IS_ERR(inode)) { + gossip_err("*** Failed to allocate pvfs2 file inode\n"); + ret = PTR_ERR(inode); + goto out; + } + + gossip_debug(GOSSIP_NAME_DEBUG, + "Assigned file inode new number of %pU\n", + get_khandle_from_ino(inode)); + + d_instantiate(dentry, inode); + unlock_new_inode(inode); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Inode (Regular File) %pU -> %s\n", + get_khandle_from_ino(inode), + dentry->d_name.name); + + SetMtimeFlag(parent); + dir->i_mtime = dir->i_ctime = current_fs_time(dir->i_sb); + mark_inode_dirty_sync(dir); + ret = 0; +out: + op_release(new_op); + gossip_debug(GOSSIP_NAME_DEBUG, "%s: returning %d\n", __func__, ret); + return ret; +} + +/* + * Attempt to resolve an object name (dentry->d_name), parent handle, and + * fsid into a handle for the object. + */ +static struct dentry *pvfs2_lookup(struct inode *dir, struct dentry *dentry, + unsigned int flags) +{ + struct pvfs2_inode_s *parent = PVFS2_I(dir); + struct pvfs2_kernel_op_s *new_op; + struct inode *inode; + struct dentry *res; + int ret = -EINVAL; + + /* + * in theory we could skip a lookup here (if the intent is to + * create) in order to avoid a potentially failed lookup, but + * leaving it in can skip a valid lookup and try to create a file + * that already exists (e.g. the vfs already handles checking for + * -EEXIST on O_EXCL opens, which is broken if we skip this lookup + * in the create path) + */ + gossip_debug(GOSSIP_NAME_DEBUG, "%s called on %s\n", + __func__, dentry->d_name.name); + + if (dentry->d_name.len > (PVFS2_NAME_LEN - 1)) + return ERR_PTR(-ENAMETOOLONG); + + new_op = op_alloc(PVFS2_VFS_OP_LOOKUP); + if (!new_op) + return ERR_PTR(-ENOMEM); + + new_op->upcall.req.lookup.sym_follow = flags & LOOKUP_FOLLOW; + + gossip_debug(GOSSIP_NAME_DEBUG, "%s:%s:%d using parent %pU\n", + __FILE__, + __func__, + __LINE__, + &parent->refn.khandle); + new_op->upcall.req.lookup.parent_refn = parent->refn; + + strncpy(new_op->upcall.req.lookup.d_name, dentry->d_name.name, + PVFS2_NAME_LEN); + + gossip_debug(GOSSIP_NAME_DEBUG, + "%s: doing lookup on %s under %pU,%d (follow=%s)\n", + __func__, + new_op->upcall.req.lookup.d_name, + &new_op->upcall.req.lookup.parent_refn.khandle, + new_op->upcall.req.lookup.parent_refn.fs_id, + ((new_op->upcall.req.lookup.sym_follow == + PVFS2_LOOKUP_LINK_FOLLOW) ? "yes" : "no")); + + ret = service_operation(new_op, __func__, get_interruptible_flag(dir)); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Lookup Got %pU, fsid %d (ret=%d)\n", + &new_op->downcall.resp.lookup.refn.khandle, + new_op->downcall.resp.lookup.refn.fs_id, + ret); + + if (ret < 0) { + if (ret == -ENOENT) { + /* + * if no inode was found, add a negative dentry to + * dcache anyway; if we don't, we don't hold expected + * lookup semantics and we most noticeably break + * during directory renames. + * + * however, if the operation failed or exited, do not + * add the dentry (e.g. in the case that a touch is + * issued on a file that already exists that was + * interrupted during this lookup -- no need to add + * another negative dentry for an existing file) + */ + + gossip_debug(GOSSIP_NAME_DEBUG, + "pvfs2_lookup: Adding *negative* dentry " + "%p for %s\n", + dentry, + dentry->d_name.name); + + d_add(dentry, NULL); + res = NULL; + goto out; + } + + /* must be a non-recoverable error */ + res = ERR_PTR(ret); + goto out; + } + + inode = pvfs2_iget(dir->i_sb, &new_op->downcall.resp.lookup.refn); + if (IS_ERR(inode)) { + gossip_debug(GOSSIP_NAME_DEBUG, + "error %ld from iget\n", PTR_ERR(inode)); + res = ERR_CAST(inode); + goto out; + } + + gossip_debug(GOSSIP_NAME_DEBUG, + "%s:%s:%d " + "Found good inode [%lu] with count [%d]\n", + __FILE__, + __func__, + __LINE__, + inode->i_ino, + (int)atomic_read(&inode->i_count)); + + /* update dentry/inode pair into dcache */ + res = d_splice_alias(inode, dentry); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Lookup success (inode ct = %d)\n", + (int)atomic_read(&inode->i_count)); +out: + op_release(new_op); + return res; +} + +/* return 0 on success; non-zero otherwise */ +static int pvfs2_unlink(struct inode *dir, struct dentry *dentry) +{ + struct inode *inode = dentry->d_inode; + struct pvfs2_inode_s *parent = PVFS2_I(dir); + struct pvfs2_kernel_op_s *new_op; + int ret; + + gossip_debug(GOSSIP_NAME_DEBUG, + "%s: called on %s\n" + " (inode %pU): Parent is %pU | fs_id %d\n", + __func__, + dentry->d_name.name, + get_khandle_from_ino(inode), + &parent->refn.khandle, + parent->refn.fs_id); + + new_op = op_alloc(PVFS2_VFS_OP_REMOVE); + if (!new_op) + return -ENOMEM; + + new_op->upcall.req.remove.parent_refn = parent->refn; + strncpy(new_op->upcall.req.remove.d_name, dentry->d_name.name, + PVFS2_NAME_LEN); + + ret = service_operation(new_op, "pvfs2_unlink", + get_interruptible_flag(inode)); + + /* when request is serviced properly, free req op struct */ + op_release(new_op); + + if (!ret) { + drop_nlink(inode); + + SetMtimeFlag(parent); + dir->i_mtime = dir->i_ctime = current_fs_time(dir->i_sb); + mark_inode_dirty_sync(dir); + } + return ret; +} + +/* + * pvfs2_link() is only implemented here to make sure that we return a + * reasonable error code (the kernel will return a misleading EPERM + * otherwise). PVFS2 does not support hard links. + */ +static int pvfs2_link(struct dentry *old_dentry, + struct inode *dir, + struct dentry *dentry) +{ + return -EOPNOTSUPP; +} + +/* + * pvfs2_mknod() is only implemented here to make sure that we return a + * reasonable error code (the kernel will return a misleading EPERM + * otherwise). PVFS2 does not support special files such as fifos or devices. + */ +static int pvfs2_mknod(struct inode *dir, + struct dentry *dentry, + umode_t mode, + dev_t rdev) +{ + return -EOPNOTSUPP; +} + +static int pvfs2_symlink(struct inode *dir, + struct dentry *dentry, + const char *symname) +{ + struct pvfs2_inode_s *parent = PVFS2_I(dir); + struct pvfs2_kernel_op_s *new_op; + struct inode *inode; + int mode = 755; + int ret; + + gossip_debug(GOSSIP_NAME_DEBUG, "%s: called\n", __func__); + + if (!symname) + return -EINVAL; + + new_op = op_alloc(PVFS2_VFS_OP_SYMLINK); + if (!new_op) + return -ENOMEM; + + new_op->upcall.req.sym.parent_refn = parent->refn; + + fill_default_sys_attrs(new_op->upcall.req.sym.attributes, + PVFS_TYPE_SYMLINK, + mode); + + strncpy(new_op->upcall.req.sym.entry_name, + dentry->d_name.name, + PVFS2_NAME_LEN); + strncpy(new_op->upcall.req.sym.target, symname, PVFS2_NAME_LEN); + + ret = service_operation(new_op, __func__, get_interruptible_flag(dir)); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Symlink Got PVFS2 handle %pU on fsid %d (ret=%d)\n", + &new_op->downcall.resp.sym.refn.khandle, + new_op->downcall.resp.sym.refn.fs_id, ret); + + if (ret < 0) { + gossip_debug(GOSSIP_NAME_DEBUG, + "%s: failed with error code %d\n", + __func__, ret); + goto out; + } + + inode = pvfs2_new_inode(dir->i_sb, dir, S_IFLNK | mode, 0, + &new_op->downcall.resp.sym.refn); + if (IS_ERR(inode)) { + gossip_err + ("*** Failed to allocate pvfs2 symlink inode\n"); + ret = PTR_ERR(inode); + goto out; + } + + gossip_debug(GOSSIP_NAME_DEBUG, + "Assigned symlink inode new number of %pU\n", + get_khandle_from_ino(inode)); + + d_instantiate(dentry, inode); + unlock_new_inode(inode); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Inode (Symlink) %pU -> %s\n", + get_khandle_from_ino(inode), + dentry->d_name.name); + + SetMtimeFlag(parent); + dir->i_mtime = dir->i_ctime = current_fs_time(dir->i_sb); + mark_inode_dirty_sync(dir); + ret = 0; +out: + op_release(new_op); + return ret; +} + +static int pvfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) +{ + struct pvfs2_inode_s *parent = PVFS2_I(dir); + struct pvfs2_kernel_op_s *new_op; + struct inode *inode; + int ret; + + new_op = op_alloc(PVFS2_VFS_OP_MKDIR); + if (!new_op) + return -ENOMEM; + + new_op->upcall.req.mkdir.parent_refn = parent->refn; + + fill_default_sys_attrs(new_op->upcall.req.mkdir.attributes, + PVFS_TYPE_DIRECTORY, mode); + + strncpy(new_op->upcall.req.mkdir.d_name, + dentry->d_name.name, PVFS2_NAME_LEN); + + ret = service_operation(new_op, __func__, get_interruptible_flag(dir)); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Mkdir Got PVFS2 handle %pU on fsid %d\n", + &new_op->downcall.resp.mkdir.refn.khandle, + new_op->downcall.resp.mkdir.refn.fs_id); + + if (ret < 0) { + gossip_debug(GOSSIP_NAME_DEBUG, + "%s: failed with error code %d\n", + __func__, ret); + goto out; + } + + inode = pvfs2_new_inode(dir->i_sb, dir, S_IFDIR | mode, 0, + &new_op->downcall.resp.mkdir.refn); + if (IS_ERR(inode)) { + gossip_err("*** Failed to allocate pvfs2 dir inode\n"); + ret = PTR_ERR(inode); + goto out; + } + + gossip_debug(GOSSIP_NAME_DEBUG, + "Assigned dir inode new number of %pU\n", + get_khandle_from_ino(inode)); + + d_instantiate(dentry, inode); + unlock_new_inode(inode); + + gossip_debug(GOSSIP_NAME_DEBUG, + "Inode (Directory) %pU -> %s\n", + get_khandle_from_ino(inode), + dentry->d_name.name); + + /* + * NOTE: we have no good way to keep nlink consistent for directories + * across clients; keep constant at 1. + */ + SetMtimeFlag(parent); + dir->i_mtime = dir->i_ctime = current_fs_time(dir->i_sb); + mark_inode_dirty_sync(dir); +out: + op_release(new_op); + return ret; +} + +static int pvfs2_rename(struct inode *old_dir, + struct dentry *old_dentry, + struct inode *new_dir, + struct dentry *new_dentry) +{ + struct pvfs2_kernel_op_s *new_op; + int ret; + + gossip_debug(GOSSIP_NAME_DEBUG, + "pvfs2_rename: called (%s/%s => %s/%s) ct=%d\n", + old_dentry->d_parent->d_name.name, + old_dentry->d_name.name, + new_dentry->d_parent->d_name.name, + new_dentry->d_name.name, + d_count(new_dentry)); + + new_op = op_alloc(PVFS2_VFS_OP_RENAME); + if (!new_op) + return -EINVAL; + + new_op->upcall.req.rename.old_parent_refn = PVFS2_I(old_dir)->refn; + new_op->upcall.req.rename.new_parent_refn = PVFS2_I(new_dir)->refn; + + strncpy(new_op->upcall.req.rename.d_old_name, + old_dentry->d_name.name, + PVFS2_NAME_LEN); + strncpy(new_op->upcall.req.rename.d_new_name, + new_dentry->d_name.name, + PVFS2_NAME_LEN); + + ret = service_operation(new_op, + "pvfs2_rename", + get_interruptible_flag(old_dentry->d_inode)); + + gossip_debug(GOSSIP_NAME_DEBUG, + "pvfs2_rename: got downcall status %d\n", + ret); + + if (new_dentry->d_inode) + new_dentry->d_inode->i_ctime = CURRENT_TIME; + + op_release(new_op); + return ret; +} + +/* PVFS2 implementation of VFS inode operations for directories */ +struct inode_operations pvfs2_dir_inode_operations = { + .lookup = pvfs2_lookup, + .get_acl = pvfs2_get_acl, + .set_acl = pvfs2_set_acl, + .create = pvfs2_create, + .link = pvfs2_link, + .unlink = pvfs2_unlink, + .symlink = pvfs2_symlink, + .mkdir = pvfs2_mkdir, + .rmdir = pvfs2_unlink, + .mknod = pvfs2_mknod, + .rename = pvfs2_rename, + .setattr = pvfs2_setattr, + .getattr = pvfs2_getattr, + .setxattr = generic_setxattr, + .getxattr = generic_getxattr, + .removexattr = generic_removexattr, + .listxattr = pvfs2_listxattr, +}; diff --git a/fs/orangefs/pvfs2-bufmap.c b/fs/orangefs/pvfs2-bufmap.c new file mode 100644 index 000000000000..aa14c37d0216 --- /dev/null +++ b/fs/orangefs/pvfs2-bufmap.c @@ -0,0 +1,970 @@ +/* + * (C) 2001 Clemson University and The University of Chicago + * + * See COPYING in top-level directory. + */ +#include "protocol.h" +#include "pvfs2-kernel.h" +#include "pvfs2-bufmap.h" + +DECLARE_WAIT_QUEUE_HEAD(pvfs2_bufmap_init_waitq); + +struct pvfs2_bufmap { + atomic_t refcnt; + + int desc_size; + int desc_shift; + int desc_count; + int total_size; + int page_count; + + struct page **page_array; + struct pvfs_bufmap_desc *desc_array; + + /* array to track usage of buffer descriptors */ + int *buffer_index_array; + spinlock_t buffer_index_lock; + + /* array to track usage of buffer descriptors for readdir */ + int readdir_index_array[PVFS2_READDIR_DEFAULT_DESC_COUNT]; + spinlock_t readdir_index_lock; +} *__pvfs2_bufmap; + +static DEFINE_SPINLOCK(pvfs2_bufmap_lock); + +static void +pvfs2_bufmap_unmap(struct pvfs2_bufmap *bufmap) +{ + int i; + + for (i = 0; i < bufmap->page_count; i++) + page_cache_release(bufmap->page_array[i]); +} + +static void +pvfs2_bufmap_free(struct pvfs2_bufmap *bufmap) +{ + kfree(bufmap->page_array); + kfree(bufmap->desc_array); + kfree(bufmap->buffer_index_array); + kfree(bufmap); +} + +struct pvfs2_bufmap *pvfs2_bufmap_ref(void) +{ + struct pvfs2_bufmap *bufmap = NULL; + + spin_lock(&pvfs2_bufmap_lock); + if (__pvfs2_bufmap) { + bufmap = __pvfs2_bufmap; + atomic_inc(&bufmap->refcnt); + } + spin_unlock(&pvfs2_bufmap_lock); + return bufmap; +} + +void pvfs2_bufmap_unref(struct pvfs2_bufmap *bufmap) +{ + if (atomic_dec_and_lock(&bufmap->refcnt, &pvfs2_bufmap_lock)) { + __pvfs2_bufmap = NULL; + spin_unlock(&pvfs2_bufmap_lock); + + pvfs2_bufmap_unmap(bufmap); + pvfs2_bufmap_free(bufmap); + } +} + +inline int pvfs_bufmap_size_query(void) +{ + struct pvfs2_bufmap *bufmap = pvfs2_bufmap_ref(); + int size = bufmap ? bufmap->desc_size : 0; + + pvfs2_bufmap_unref(bufmap); + return size; +} + +inline int pvfs_bufmap_shift_query(void) +{ + struct pvfs2_bufmap *bufmap = pvfs2_bufmap_ref(); + int shift = bufmap ? bufmap->desc_shift : 0; + + pvfs2_bufmap_unref(bufmap); + return shift; +} + +static DECLARE_WAIT_QUEUE_HEAD(bufmap_waitq); +static DECLARE_WAIT_QUEUE_HEAD(readdir_waitq); + +/* + * get_bufmap_init + * + * If bufmap_init is 1, then the shared memory system, including the + * buffer_index_array, is available. Otherwise, it is not. + * + * returns the value of bufmap_init + */ +int get_bufmap_init(void) +{ + return __pvfs2_bufmap ? 1 : 0; +} + + +static struct pvfs2_bufmap * +pvfs2_bufmap_alloc(struct PVFS_dev_map_desc *user_desc) +{ + struct pvfs2_bufmap *bufmap; + + bufmap = kzalloc(sizeof(*bufmap), GFP_KERNEL); + if (!bufmap) + goto out; + + atomic_set(&bufmap->refcnt, 1); + bufmap->total_size = user_desc->total_size; + bufmap->desc_count = user_desc->count; + bufmap->desc_size = user_desc->size; + bufmap->desc_shift = ilog2(bufmap->desc_size); + + spin_lock_init(&bufmap->buffer_index_lock); + bufmap->buffer_index_array = + kcalloc(bufmap->desc_count, sizeof(int), GFP_KERNEL); + if (!bufmap->buffer_index_array) { + gossip_err("pvfs2: could not allocate %d buffer indices\n", + bufmap->desc_count); + goto out_free_bufmap; + } + spin_lock_init(&bufmap->readdir_index_lock); + + bufmap->desc_array = + kcalloc(bufmap->desc_count, sizeof(struct pvfs_bufmap_desc), + GFP_KERNEL); + if (!bufmap->desc_array) { + gossip_err("pvfs2: could not allocate %d descriptors\n", + bufmap->desc_count); + goto out_free_index_array; + } + + bufmap->page_count = bufmap->total_size / PAGE_SIZE; + + /* allocate storage to track our page mappings */ + bufmap->page_array = + kcalloc(bufmap->page_count, sizeof(struct page *), GFP_KERNEL); + if (!bufmap->page_array) + goto out_free_desc_array; + + return bufmap; + +out_free_desc_array: + kfree(bufmap->desc_array); +out_free_index_array: + kfree(bufmap->buffer_index_array); +out_free_bufmap: + kfree(bufmap); +out: + return NULL; +} + +static int +pvfs2_bufmap_map(struct pvfs2_bufmap *bufmap, + struct PVFS_dev_map_desc *user_desc) +{ + int pages_per_desc = bufmap->desc_size / PAGE_SIZE; + int offset = 0, ret, i; + + /* map the pages */ + down_write(¤t->mm->mmap_sem); + ret = get_user_pages(current, + current->mm, + (unsigned long)user_desc->ptr, + bufmap->page_count, + 1, + 0, + bufmap->page_array, + NULL); + up_write(¤t->mm->mmap_sem); + + if (ret < 0) + return ret; + + if (ret != bufmap->page_count) { + gossip_err("pvfs2 error: asked for %d pages, only got %d.\n", + bufmap->page_count, ret); + + for (i = 0; i < ret; i++) { + SetPageError(bufmap->page_array[i]); + page_cache_release(bufmap->page_array[i]); + } + return -ENOMEM; + } + + /* + * ideally we want to get kernel space pointers for each page, but + * we can't kmap that many pages at once if highmem is being used. + * so instead, we just kmap/kunmap the page address each time the + * kaddr is needed. + */ + for (i = 0; i < bufmap->page_count; i++) + flush_dcache_page(bufmap->page_array[i]); + + /* build a list of available descriptors */ + for (offset = 0, i = 0; i < bufmap->desc_count; i++) { + bufmap->desc_array[i].page_array = &bufmap->page_array[offset]; + bufmap->desc_array[i].array_count = pages_per_desc; + bufmap->desc_array[i].uaddr = + (user_desc->ptr + (i * pages_per_desc * PAGE_SIZE)); + offset += pages_per_desc; + } + + return 0; +} + +/* + * pvfs_bufmap_initialize() + * + * initializes the mapped buffer interface + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_initialize(struct PVFS_dev_map_desc *user_desc) +{ + struct pvfs2_bufmap *bufmap; + int ret = -EINVAL; + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_initialize: called (ptr (" + "%p) sz (%d) cnt(%d).\n", + user_desc->ptr, + user_desc->size, + user_desc->count); + + /* + * sanity check alignment and size of buffer that caller wants to + * work with + */ + if (PAGE_ALIGN((unsigned long)user_desc->ptr) != + (unsigned long)user_desc->ptr) { + gossip_err("pvfs2 error: memory alignment (front). %p\n", + user_desc->ptr); + goto out; + } + + if (PAGE_ALIGN(((unsigned long)user_desc->ptr + user_desc->total_size)) + != (unsigned long)(user_desc->ptr + user_desc->total_size)) { + gossip_err("pvfs2 error: memory alignment (back).(%p + %d)\n", + user_desc->ptr, + user_desc->total_size); + goto out; + } + + if (user_desc->total_size != (user_desc->size * user_desc->count)) { + gossip_err("pvfs2 error: user provided an oddly sized buffer: (%d, %d, %d)\n", + user_desc->total_size, + user_desc->size, + user_desc->count); + goto out; + } + + if ((user_desc->size % PAGE_SIZE) != 0) { + gossip_err("pvfs2 error: bufmap size not page size divisible (%d).\n", + user_desc->size); + goto out; + } + + ret = -ENOMEM; + bufmap = pvfs2_bufmap_alloc(user_desc); + if (!bufmap) + goto out; + + ret = pvfs2_bufmap_map(bufmap, user_desc); + if (ret) + goto out_free_bufmap; + + + spin_lock(&pvfs2_bufmap_lock); + if (__pvfs2_bufmap) { + spin_unlock(&pvfs2_bufmap_lock); + gossip_err("pvfs2: error: bufmap already initialized.\n"); + ret = -EALREADY; + goto out_unmap_bufmap; + } + __pvfs2_bufmap = bufmap; + spin_unlock(&pvfs2_bufmap_lock); + + /* + * If there are operations in pvfs2_bufmap_init_waitq, wake them up. + * This scenario occurs when the client-core is restarted and I/O + * requests in the in-progress or waiting tables are restarted. I/O + * requests cannot be restarted until the shared memory system is + * completely re-initialized, so we put the I/O requests in this + * waitq until initialization has completed. NOTE: the I/O requests + * are also on a timer, so they don't wait forever just in case the + * client-core doesn't come back up. + */ + wake_up_interruptible(&pvfs2_bufmap_init_waitq); + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_initialize: exiting normally\n"); + return 0; + +out_unmap_bufmap: + pvfs2_bufmap_unmap(bufmap); +out_free_bufmap: + pvfs2_bufmap_free(bufmap); +out: + return ret; +} + +/* + * pvfs_bufmap_finalize() + * + * shuts down the mapped buffer interface and releases any resources + * associated with it + * + * no return value + */ +void pvfs_bufmap_finalize(void) +{ + gossip_debug(GOSSIP_BUFMAP_DEBUG, "pvfs2_bufmap_finalize: called\n"); + BUG_ON(!__pvfs2_bufmap); + pvfs2_bufmap_unref(__pvfs2_bufmap); + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs2_bufmap_finalize: exiting normally\n"); +} + +struct slot_args { + int slot_count; + int *slot_array; + spinlock_t *slot_lock; + wait_queue_head_t *slot_wq; +}; + +static int wait_for_a_slot(struct slot_args *slargs, int *buffer_index) +{ + int ret = -1; + int i = 0; + DECLARE_WAITQUEUE(my_wait, current); + + + add_wait_queue_exclusive(slargs->slot_wq, &my_wait); + + while (1) { + set_current_state(TASK_INTERRUPTIBLE); + + /* + * check for available desc, slot_lock is the appropriate + * index_lock + */ + spin_lock(slargs->slot_lock); + for (i = 0; i < slargs->slot_count; i++) + if (slargs->slot_array[i] == 0) { + slargs->slot_array[i] = 1; + *buffer_index = i; + ret = 0; + break; + } + spin_unlock(slargs->slot_lock); + + /* if we acquired a buffer, then break out of while */ + if (ret == 0) + break; + + if (!signal_pending(current)) { + int timeout = + MSECS_TO_JIFFIES(1000 * slot_timeout_secs); + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "[BUFMAP]: waiting %d " + "seconds for a slot\n", + slot_timeout_secs); + if (!schedule_timeout(timeout)) { + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "*** wait_for_a_slot timed out\n"); + ret = -ETIMEDOUT; + break; + } + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "[BUFMAP]: woken up by a slot becoming available.\n"); + continue; + } + + gossip_debug(GOSSIP_BUFMAP_DEBUG, "pvfs2: %s interrupted.\n", + __func__); + ret = -EINTR; + break; + } + + set_current_state(TASK_RUNNING); + remove_wait_queue(slargs->slot_wq, &my_wait); + return ret; +} + +static void put_back_slot(struct slot_args *slargs, int buffer_index) +{ + /* slot_lock is the appropriate index_lock */ + spin_lock(slargs->slot_lock); + if (buffer_index < 0 || buffer_index >= slargs->slot_count) { + spin_unlock(slargs->slot_lock); + return; + } + + /* put the desc back on the queue */ + slargs->slot_array[buffer_index] = 0; + spin_unlock(slargs->slot_lock); + + /* wake up anyone who may be sleeping on the queue */ + wake_up_interruptible(slargs->slot_wq); +} + +/* + * pvfs_bufmap_get() + * + * gets a free mapped buffer descriptor, will sleep until one becomes + * available if necessary + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_get(struct pvfs2_bufmap **mapp, int *buffer_index) +{ + struct pvfs2_bufmap *bufmap = pvfs2_bufmap_ref(); + struct slot_args slargs; + int ret; + + if (!bufmap) { + gossip_err("pvfs2: please confirm that pvfs2-client daemon is running.\n"); + return -EIO; + } + + slargs.slot_count = bufmap->desc_count; + slargs.slot_array = bufmap->buffer_index_array; + slargs.slot_lock = &bufmap->buffer_index_lock; + slargs.slot_wq = &bufmap_waitq; + ret = wait_for_a_slot(&slargs, buffer_index); + if (ret) + pvfs2_bufmap_unref(bufmap); + *mapp = bufmap; + return ret; +} + +/* + * pvfs_bufmap_put() + * + * returns a mapped buffer descriptor to the collection + * + * no return value + */ +void pvfs_bufmap_put(struct pvfs2_bufmap *bufmap, int buffer_index) +{ + struct slot_args slargs; + + slargs.slot_count = bufmap->desc_count; + slargs.slot_array = bufmap->buffer_index_array; + slargs.slot_lock = &bufmap->buffer_index_lock; + slargs.slot_wq = &bufmap_waitq; + put_back_slot(&slargs, buffer_index); + pvfs2_bufmap_unref(bufmap); +} + +/* + * readdir_index_get() + * + * gets a free descriptor, will sleep until one becomes + * available if necessary. + * Although the readdir buffers are not mapped into kernel space + * we could do that at a later point of time. Regardless, these + * indices are used by the client-core. + * + * returns 0 on success, -errno on failure + */ +int readdir_index_get(struct pvfs2_bufmap **mapp, int *buffer_index) +{ + struct pvfs2_bufmap *bufmap = pvfs2_bufmap_ref(); + struct slot_args slargs; + int ret; + + if (!bufmap) { + gossip_err("pvfs2: please confirm that pvfs2-client daemon is running.\n"); + return -EIO; + } + + slargs.slot_count = PVFS2_READDIR_DEFAULT_DESC_COUNT; + slargs.slot_array = bufmap->readdir_index_array; + slargs.slot_lock = &bufmap->readdir_index_lock; + slargs.slot_wq = &readdir_waitq; + ret = wait_for_a_slot(&slargs, buffer_index); + if (ret) + pvfs2_bufmap_unref(bufmap); + *mapp = bufmap; + return ret; +} + +void readdir_index_put(struct pvfs2_bufmap *bufmap, int buffer_index) +{ + struct slot_args slargs; + + slargs.slot_count = PVFS2_READDIR_DEFAULT_DESC_COUNT; + slargs.slot_array = bufmap->readdir_index_array; + slargs.slot_lock = &bufmap->readdir_index_lock; + slargs.slot_wq = &readdir_waitq; + put_back_slot(&slargs, buffer_index); + pvfs2_bufmap_unref(bufmap); +} + +/* + * pvfs_bufmap_copy_iovec_from_user() + * + * copies data from several user space address's in an iovec + * to a mapped buffer + * + * Note that the mapped buffer is a series of pages and therefore + * the copies have to be split by PAGE_SIZE bytes at a time. + * Note that this routine checks that summation of iov_len + * across all the elements of iov is equal to size. + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_copy_iovec_from_user(struct pvfs2_bufmap *bufmap, + int buffer_index, + const struct iovec *iov, + unsigned long nr_segs, + size_t size) +{ + size_t ret = 0; + size_t amt_copied = 0; + size_t cur_copy_size = 0; + unsigned int to_page_offset = 0; + unsigned int to_page_index = 0; + void *to_kaddr = NULL; + void __user *from_addr = NULL; + struct iovec *copied_iovec = NULL; + struct pvfs_bufmap_desc *to; + unsigned int seg; + char *tmp_printer = NULL; + int tmp_int = 0; + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_copy_iovec_from_user: index %d, " + "size %zd\n", + buffer_index, + size); + + to = &bufmap->desc_array[buffer_index]; + + /* + * copy the passed in iovec so that we can change some of its fields + */ + copied_iovec = kmalloc_array(nr_segs, + sizeof(*copied_iovec), + PVFS2_BUFMAP_GFP_FLAGS); + if (copied_iovec == NULL) + return -ENOMEM; + + memcpy(copied_iovec, iov, nr_segs * sizeof(*copied_iovec)); + /* + * Go through each segment in the iovec and make sure that + * the summation of iov_len matches the given size. + */ + for (seg = 0, amt_copied = 0; seg < nr_segs; seg++) + amt_copied += copied_iovec[seg].iov_len; + if (amt_copied != size) { + gossip_err( + "pvfs2_bufmap_copy_iovec_from_user: computed total (" + "%zd) is not equal to (%zd)\n", + amt_copied, + size); + kfree(copied_iovec); + return -EINVAL; + } + + to_page_index = 0; + to_page_offset = 0; + amt_copied = 0; + seg = 0; + /* + * Go through each segment in the iovec and copy its + * buffer into the mapped buffer one page at a time though + */ + while (amt_copied < size) { + struct iovec *iv = &copied_iovec[seg]; + int inc_to_page_index; + + if (iv->iov_len < (PAGE_SIZE - to_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + from_addr = iv->iov_base; + inc_to_page_index = 0; + } else if (iv->iov_len == (PAGE_SIZE - to_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + from_addr = iv->iov_base; + inc_to_page_index = 1; + } else { + cur_copy_size = + PVFS_util_min(PAGE_SIZE - to_page_offset, + size - amt_copied); + from_addr = iv->iov_base; + iv->iov_base += cur_copy_size; + iv->iov_len -= cur_copy_size; + inc_to_page_index = 1; + } + to_kaddr = pvfs2_kmap(to->page_array[to_page_index]); + ret = + copy_from_user(to_kaddr + to_page_offset, + from_addr, + cur_copy_size); + if (!PageReserved(to->page_array[to_page_index])) + SetPageDirty(to->page_array[to_page_index]); + + if (!tmp_printer) { + tmp_printer = (char *)(to_kaddr + to_page_offset); + tmp_int += tmp_printer[0]; + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "First character (integer value) in pvfs_bufmap_copy_from_user: %d\n", + tmp_int); + } + + pvfs2_kunmap(to->page_array[to_page_index]); + if (ret) { + gossip_err("Failed to copy data from user space\n"); + kfree(copied_iovec); + return -EFAULT; + } + + amt_copied += cur_copy_size; + if (inc_to_page_index) { + to_page_offset = 0; + to_page_index++; + } else { + to_page_offset += cur_copy_size; + } + } + kfree(copied_iovec); + return 0; +} + +/* + * pvfs_bufmap_copy_iovec_from_kernel() + * + * copies data from several kernel space address's in an iovec + * to a mapped buffer + * + * Note that the mapped buffer is a series of pages and therefore + * the copies have to be split by PAGE_SIZE bytes at a time. + * Note that this routine checks that summation of iov_len + * across all the elements of iov is equal to size. + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_copy_iovec_from_kernel(struct pvfs2_bufmap *bufmap, + int buffer_index, const struct iovec *iov, + unsigned long nr_segs, size_t size) +{ + size_t amt_copied = 0; + size_t cur_copy_size = 0; + int to_page_index = 0; + void *to_kaddr = NULL; + void *from_kaddr = NULL; + struct iovec *copied_iovec = NULL; + struct pvfs_bufmap_desc *to; + unsigned int seg; + unsigned to_page_offset = 0; + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_copy_iovec_from_kernel: index %d, " + "size %zd\n", + buffer_index, + size); + + to = &bufmap->desc_array[buffer_index]; + /* + * copy the passed in iovec so that we can change some of its fields + */ + copied_iovec = kmalloc_array(nr_segs, + sizeof(*copied_iovec), + PVFS2_BUFMAP_GFP_FLAGS); + if (copied_iovec == NULL) + return -ENOMEM; + + memcpy(copied_iovec, iov, nr_segs * sizeof(*copied_iovec)); + /* + * Go through each segment in the iovec and make sure that + * the summation of iov_len matches the given size. + */ + for (seg = 0, amt_copied = 0; seg < nr_segs; seg++) + amt_copied += copied_iovec[seg].iov_len; + if (amt_copied != size) { + gossip_err("pvfs2_bufmap_copy_iovec_from_kernel: computed total(%zd) is not equal to (%zd)\n", + amt_copied, + size); + kfree(copied_iovec); + return -EINVAL; + } + + to_page_index = 0; + amt_copied = 0; + seg = 0; + to_page_offset = 0; + /* + * Go through each segment in the iovec and copy its + * buffer into the mapped buffer one page at a time though + */ + while (amt_copied < size) { + struct iovec *iv = &copied_iovec[seg]; + int inc_to_page_index; + + if (iv->iov_len < (PAGE_SIZE - to_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + from_kaddr = iv->iov_base; + inc_to_page_index = 0; + } else if (iv->iov_len == (PAGE_SIZE - to_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + from_kaddr = iv->iov_base; + inc_to_page_index = 1; + } else { + cur_copy_size = + PVFS_util_min(PAGE_SIZE - to_page_offset, + size - amt_copied); + from_kaddr = iv->iov_base; + iv->iov_base += cur_copy_size; + iv->iov_len -= cur_copy_size; + inc_to_page_index = 1; + } + to_kaddr = pvfs2_kmap(to->page_array[to_page_index]); + memcpy(to_kaddr + to_page_offset, from_kaddr, cur_copy_size); + if (!PageReserved(to->page_array[to_page_index])) + SetPageDirty(to->page_array[to_page_index]); + pvfs2_kunmap(to->page_array[to_page_index]); + amt_copied += cur_copy_size; + if (inc_to_page_index) { + to_page_offset = 0; + to_page_index++; + } else { + to_page_offset += cur_copy_size; + } + } + kfree(copied_iovec); + return 0; +} + +/* + * pvfs_bufmap_copy_to_user_iovec() + * + * copies data to several user space address's in an iovec + * from a mapped buffer + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_copy_to_user_iovec(struct pvfs2_bufmap *bufmap, + int buffer_index, const struct iovec *iov, + unsigned long nr_segs, size_t size) +{ + size_t ret = 0; + size_t amt_copied = 0; + size_t cur_copy_size = 0; + int from_page_index = 0; + void *from_kaddr = NULL; + void __user *to_addr = NULL; + struct iovec *copied_iovec = NULL; + struct pvfs_bufmap_desc *from; + unsigned int seg; + unsigned from_page_offset = 0; + char *tmp_printer = NULL; + int tmp_int = 0; + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_copy_to_user_iovec: index %d, size %zd\n", + buffer_index, + size); + + from = &bufmap->desc_array[buffer_index]; + /* + * copy the passed in iovec so that we can change some of its fields + */ + copied_iovec = kmalloc_array(nr_segs, + sizeof(*copied_iovec), + PVFS2_BUFMAP_GFP_FLAGS); + if (copied_iovec == NULL) + return -ENOMEM; + + memcpy(copied_iovec, iov, nr_segs * sizeof(*copied_iovec)); + /* + * Go through each segment in the iovec and make sure that + * the summation of iov_len is greater than the given size. + */ + for (seg = 0, amt_copied = 0; seg < nr_segs; seg++) + amt_copied += copied_iovec[seg].iov_len; + if (amt_copied < size) { + gossip_err("pvfs2_bufmap_copy_to_user_iovec: computed total (%zd) is less than (%zd)\n", + amt_copied, + size); + kfree(copied_iovec); + return -EINVAL; + } + + from_page_index = 0; + amt_copied = 0; + seg = 0; + from_page_offset = 0; + /* + * Go through each segment in the iovec and copy from the mapper buffer, + * but make sure that we do so one page at a time. + */ + while (amt_copied < size) { + struct iovec *iv = &copied_iovec[seg]; + int inc_from_page_index; + + if (iv->iov_len < (PAGE_SIZE - from_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + to_addr = iv->iov_base; + inc_from_page_index = 0; + } else if (iv->iov_len == (PAGE_SIZE - from_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + to_addr = iv->iov_base; + inc_from_page_index = 1; + } else { + cur_copy_size = + PVFS_util_min(PAGE_SIZE - from_page_offset, + size - amt_copied); + to_addr = iv->iov_base; + iv->iov_base += cur_copy_size; + iv->iov_len -= cur_copy_size; + inc_from_page_index = 1; + } + from_kaddr = pvfs2_kmap(from->page_array[from_page_index]); + if (!tmp_printer) { + tmp_printer = (char *)(from_kaddr + from_page_offset); + tmp_int += tmp_printer[0]; + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "First character (integer value) in pvfs_bufmap_copy_to_user_iovec: %d\n", + tmp_int); + } + ret = + copy_to_user(to_addr, + from_kaddr + from_page_offset, + cur_copy_size); + pvfs2_kunmap(from->page_array[from_page_index]); + if (ret) { + gossip_err("Failed to copy data to user space\n"); + kfree(copied_iovec); + return -EFAULT; + } + + amt_copied += cur_copy_size; + if (inc_from_page_index) { + from_page_offset = 0; + from_page_index++; + } else { + from_page_offset += cur_copy_size; + } + } + kfree(copied_iovec); + return 0; +} + +/* + * pvfs_bufmap_copy_to_kernel_iovec() + * + * copies data to several kernel space address's in an iovec + * from a mapped buffer + * + * returns 0 on success, -errno on failure + */ +int pvfs_bufmap_copy_to_kernel_iovec(struct pvfs2_bufmap *bufmap, + int buffer_index, const struct iovec *iov, + unsigned long nr_segs, size_t size) +{ + size_t amt_copied = 0; + size_t cur_copy_size = 0; + int from_page_index = 0; + void *from_kaddr = NULL; + void *to_kaddr = NULL; + struct iovec *copied_iovec = NULL; + struct pvfs_bufmap_desc *from; + unsigned int seg; + unsigned int from_page_offset = 0; + + gossip_debug(GOSSIP_BUFMAP_DEBUG, + "pvfs_bufmap_copy_to_kernel_iovec: index %d, size %zd\n", + buffer_index, + size); + + from = &bufmap->desc_array[buffer_index]; + /* + * copy the passed in iovec so that we can change some of its fields + */ + copied_iovec = kmalloc_array(nr_segs, + sizeof(*copied_iovec), + PVFS2_BUFMAP_GFP_FLAGS); + if (copied_iovec == NULL) + return -ENOMEM; + + memcpy(copied_iovec, iov, nr_segs * sizeof(*copied_iovec)); + /* + * Go through each segment in the iovec and make sure that + * the summation of iov_len is greater than the given size. + */ + for (seg = 0, amt_copied = 0; seg < nr_segs; seg++) + amt_copied += copied_iovec[seg].iov_len; + + if (amt_copied < size) { + gossip_err("pvfs2_bufmap_copy_to_kernel_iovec: computed total (%zd) is less than (%zd)\n", + amt_copied, + size); + kfree(copied_iovec); + return -EINVAL; + } + + from_page_index = 0; + amt_copied = 0; + seg = 0; + from_page_offset = 0; + /* + * Go through each segment in the iovec and copy from the mapper buffer, + * but make sure that we do so one page at a time. + */ + while (amt_copied < size) { + struct iovec *iv = &copied_iovec[seg]; + int inc_from_page_index; + + if (iv->iov_len < (PAGE_SIZE - from_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + to_kaddr = iv->iov_base; + inc_from_page_index = 0; + } else if (iv->iov_len == (PAGE_SIZE - from_page_offset)) { + cur_copy_size = + PVFS_util_min(iv->iov_len, size - amt_copied); + seg++; + to_kaddr = iv->iov_base; + inc_from_page_index = 1; + } else { + cur_copy_size = + PVFS_util_min(PAGE_SIZE - from_page_offset, + size - amt_copied); + to_kaddr = iv->iov_base; + iv->iov_base += cur_copy_size; + iv->iov_len -= cur_copy_size; + inc_from_page_index = 1; + } + from_kaddr = pvfs2_kmap(from->page_array[from_page_index]); + memcpy(to_kaddr, from_kaddr + from_page_offset, cur_copy_size); + pvfs2_kunmap(from->page_array[from_page_index]); + amt_copied += cur_copy_size; + if (inc_from_page_index) { + from_page_offset = 0; + from_page_index++; + } else { + from_page_offset += cur_copy_size; + } + } + kfree(copied_iovec); + return 0; +} diff --git a/fs/orangefs/pvfs2-cache.c b/fs/orangefs/pvfs2-cache.c new file mode 100644 index 000000000000..15251884ba4a --- /dev/null +++ b/fs/orangefs/pvfs2-cache.c @@ -0,0 +1,260 @@ +/* + * (C) 2001 Clemson University and The University of Chicago + * + * See COPYING in top-level directory. + */ + +#include "protocol.h" +#include "pvfs2-kernel.h" + +/* tags assigned to kernel upcall operations */ +static __u64 next_tag_value; +static DEFINE_SPINLOCK(next_tag_value_lock); + +/* the pvfs2 memory caches */ + +/* a cache for pvfs2 upcall/downcall operations */ +static struct kmem_cache *op_cache; + +/* a cache for device (/dev/pvfs2-req) communication */ +static struct kmem_cache *dev_req_cache; + +/* a cache for pvfs2_kiocb objects (i.e pvfs2 iocb structures ) */ +static struct kmem_cache *pvfs2_kiocb_cache; + +int op_cache_initialize(void) +{ + op_cache = kmem_cache_create("pvfs2_op_cache", + sizeof(struct pvfs2_kernel_op_s), + 0, + PVFS2_CACHE_CREATE_FLAGS, + NULL); + + if (!op_cache) { + gossip_err("Cannot create pvfs2_op_cache\n"); + return -ENOMEM; + } + + /* initialize our atomic tag counter */ + spin_lock(&next_tag_value_lock); + next_tag_value = 100; + spin_unlock(&next_tag_value_lock); + return 0; +} + +int op_cache_finalize(void) +{ + kmem_cache_destroy(op_cache); + return 0; +} + +char *get_opname_string(struct pvfs2_kernel_op_s *new_op) +{ + if (new_op) { + __s32 type = new_op->upcall.type; + + if (type == PVFS2_VFS_OP_FILE_IO) + return "OP_FILE_IO"; + else if (type == PVFS2_VFS_OP_LOOKUP) + return "OP_LOOKUP"; + else if (type == PVFS2_VFS_OP_CREATE) + return "OP_CREATE"; + else if (type == PVFS2_VFS_OP_GETATTR) + return "OP_GETATTR"; + else if (type == PVFS2_VFS_OP_REMOVE) + return "OP_REMOVE"; + else if (type == PVFS2_VFS_OP_MKDIR) + return "OP_MKDIR"; + else if (type == PVFS2_VFS_OP_READDIR) + return "OP_READDIR"; + else if (type == PVFS2_VFS_OP_READDIRPLUS) + return "OP_READDIRPLUS"; + else if (type == PVFS2_VFS_OP_SETATTR) + return "OP_SETATTR"; + else if (type == PVFS2_VFS_OP_SYMLINK) + return "OP_SYMLINK"; + else if (type == PVFS2_VFS_OP_RENAME) + return "OP_RENAME"; + else if (type == PVFS2_VFS_OP_STATFS) + return "OP_STATFS"; + else if (type == PVFS2_VFS_OP_TRUNCATE) + return "OP_TRUNCATE"; + else if (type == PVFS2_VFS_OP_MMAP_RA_FLUSH) + return "OP_MMAP_RA_FLUSH"; + else if (type == PVFS2_VFS_OP_FS_MOUNT) + return "OP_FS_MOUNT"; + else if (type == PVFS2_VFS_OP_FS_UMOUNT) + return "OP_FS_UMOUNT"; + else if (type == PVFS2_VFS_OP_GETXATTR) + return "OP_GETXATTR"; + else if (type == PVFS2_VFS_OP_SETXATTR) + return "OP_SETXATTR"; + else if (type == PVFS2_VFS_OP_LISTXATTR) + return "OP_LISTXATTR"; + else if (type == PVFS2_VFS_OP_REMOVEXATTR) + return "OP_REMOVEXATTR"; + else if (type == PVFS2_VFS_OP_PARAM) + return "OP_PARAM"; + else if (type == PVFS2_VFS_OP_PERF_COUNT) + return "OP_PERF_COUNT"; + else if (type == PVFS2_VFS_OP_CANCEL) + return "OP_CANCEL"; + else if (type == PVFS2_VFS_OP_FSYNC) + return "OP_FSYNC"; + else if (type == PVFS2_VFS_OP_FSKEY) + return "OP_FSKEY"; + else if (type == PVFS2_VFS_OP_FILE_IOX) + return "OP_FILE_IOX"; + } + return "OP_UNKNOWN?"; +} + +static struct pvfs2_kernel_op_s *op_alloc_common(__s32 op_linger, __s32 type) +{ + struct pvfs2_kernel_op_s *new_op = NULL; + + new_op = kmem_cache_alloc(op_cache, PVFS2_CACHE_ALLOC_FLAGS); + if (new_op) { + memset(new_op, 0, sizeof(struct pvfs2_kernel_op_s)); + + INIT_LIST_HEAD(&new_op->list); + spin_lock_init(&new_op->lock); + init_waitqueue_head(&new_op->waitq); + + init_waitqueue_head(&new_op->io_completion_waitq); + atomic_set(&new_op->aio_ref_count, 0); + + pvfs2_op_initialize(new_op); + + /* initialize the op specific tag and upcall credentials */ + spin_lock(&next_tag_value_lock); + new_op->tag = next_tag_value++; + if (next_tag_value == 0) + next_tag_value = 100; + spin_unlock(&next_tag_value_lock); + new_op->upcall.type = type; + new_op->attempts = 0; + gossip_debug(GOSSIP_CACHE_DEBUG, + "Alloced OP (%p: %llu %s)\n", + new_op, + llu(new_op->tag), + get_opname_string(new_op)); + + new_op->upcall.uid = from_kuid(current_user_ns(), + current_fsuid()); + + new_op->upcall.gid = from_kgid(current_user_ns(), + current_fsgid()); + + new_op->op_linger = new_op->op_linger_tmp = op_linger; + } else { + gossip_err("op_alloc: kmem_cache_alloc failed!\n"); + } + return new_op; +} + +struct pvfs2_kernel_op_s *op_alloc(__s32 type) +{ + return op_alloc_common(1, type); +} + +struct pvfs2_kernel_op_s *op_alloc_trailer(__s32 type) +{ + return op_alloc_common(2, type); +} + +void op_release(struct pvfs2_kernel_op_s *pvfs2_op) +{ + if (pvfs2_op) { + gossip_debug(GOSSIP_CACHE_DEBUG, + "Releasing OP (%p: %llu)\n", + pvfs2_op, + llu(pvfs2_op->tag)); + pvfs2_op_initialize(pvfs2_op); + kmem_cache_free(op_cache, pvfs2_op); + } else { + gossip_err("NULL pointer in op_release\n"); + } +} + +int dev_req_cache_initialize(void) +{ + dev_req_cache = kmem_cache_create("pvfs2_devreqcache", + MAX_ALIGNED_DEV_REQ_DOWNSIZE, + 0, + PVFS2_CACHE_CREATE_FLAGS, + NULL); + + if (!dev_req_cache) { + gossip_err("Cannot create pvfs2_dev_req_cache\n"); + return -ENOMEM; + } + return 0; +} + +int dev_req_cache_finalize(void) +{ + kmem_cache_destroy(dev_req_cache); + return 0; +} + +void *dev_req_alloc(void) +{ + void *buffer; + + buffer = kmem_cache_alloc(dev_req_cache, PVFS2_CACHE_ALLOC_FLAGS); + if (buffer == NULL) + gossip_err("Failed to allocate from dev_req_cache\n"); + else + memset(buffer, 0, sizeof(MAX_ALIGNED_DEV_REQ_DOWNSIZE)); + return buffer; +} + +void dev_req_release(void *buffer) +{ + if (buffer) + kmem_cache_free(dev_req_cache, buffer); + else + gossip_err("NULL pointer passed to dev_req_release\n"); +} + +int kiocb_cache_initialize(void) +{ + pvfs2_kiocb_cache = kmem_cache_create("pvfs2_kiocbcache", + sizeof(struct pvfs2_kiocb_s), + 0, + PVFS2_CACHE_CREATE_FLAGS, + NULL); + + if (!pvfs2_kiocb_cache) { + gossip_err("Cannot create pvfs2_kiocb_cache!\n"); + return -ENOMEM; + } + return 0; +} + +int kiocb_cache_finalize(void) +{ + kmem_cache_destroy(pvfs2_kiocb_cache); + return 0; +} + +struct pvfs2_kiocb_s *kiocb_alloc(void) +{ + struct pvfs2_kiocb_s *x = NULL; + + x = kmem_cache_alloc(pvfs2_kiocb_cache, PVFS2_CACHE_ALLOC_FLAGS); + if (x == NULL) + gossip_err("kiocb_alloc: kmem_cache_alloc failed!\n"); + else + memset(x, 0, sizeof(struct pvfs2_kiocb_s)); + return x; +} + +void kiocb_release(struct pvfs2_kiocb_s *x) +{ + if (x) + kmem_cache_free(pvfs2_kiocb_cache, x); + else + gossip_err("kiocb_release: kmem_cache_free NULL pointer!\n"); +} diff --git a/fs/orangefs/pvfs2-debugfs.c b/fs/orangefs/pvfs2-debugfs.c new file mode 100644 index 000000000000..8d118da9b88f --- /dev/null +++ b/fs/orangefs/pvfs2-debugfs.c @@ -0,0 +1,458 @@ +/* + * What: /sys/kernel/debug/orangefs/debug-help + * Date: June 2015 + * Contact: Mike Marshall + * Description: + * List of client and kernel debug keywords. + * + * + * What: /sys/kernel/debug/orangefs/client-debug + * Date: June 2015 + * Contact: Mike Marshall + * Description: + * Debug setting for "the client", the userspace + * helper for the kernel module. + * + * + * What: /sys/kernel/debug/orangefs/kernel-debug + * Date: June 2015 + * Contact: Mike Marshall + * Description: + * Debug setting for the orangefs kernel module. + * + * Any of the keywords, or comma-separated lists + * of keywords, from debug-help can be catted to + * client-debug or kernel-debug. + * + * "none", "all" and "verbose" are special keywords + * for client-debug. Setting client-debug to "all" + * is kind of like trying to drink water from a + * fire hose, "verbose" triggers most of the same + * output except for the constant flow of output + * from the main wait loop. + * + * "none" and "all" are similar settings for kernel-debug + * no need for a "verbose". + */ +#include +#include + +#include + +#include "pvfs2-debugfs.h" +#include "protocol.h" +#include "pvfs2-kernel.h" + +static int orangefs_debug_disabled = 1; + +static int orangefs_debug_help_open(struct inode *, struct file *); + +const struct file_operations debug_help_fops = { + .open = orangefs_debug_help_open, + .read = seq_read, + .release = seq_release, + .llseek = seq_lseek, +}; + +static void *help_start(struct seq_file *, loff_t *); +static void *help_next(struct seq_file *, void *, loff_t *); +static void help_stop(struct seq_file *, void *); +static int help_show(struct seq_file *, void *); + +static const struct seq_operations help_debug_ops = { + .start = help_start, + .next = help_next, + .stop = help_stop, + .show = help_show, +}; + +/* + * Used to protect data in ORANGEFS_KMOD_DEBUG_FILE and + * ORANGEFS_KMOD_DEBUG_FILE. + */ +DEFINE_MUTEX(orangefs_debug_lock); + +int orangefs_debug_open(struct inode *, struct file *); + +static ssize_t orangefs_debug_read(struct file *, + char __user *, + size_t, + loff_t *); + +static ssize_t orangefs_debug_write(struct file *, + const char __user *, + size_t, + loff_t *); + +static const struct file_operations kernel_debug_fops = { + .open = orangefs_debug_open, + .read = orangefs_debug_read, + .write = orangefs_debug_write, + .llseek = generic_file_llseek, +}; + +/* + * initialize kmod debug operations, create orangefs debugfs dir and + * ORANGEFS_KMOD_DEBUG_HELP_FILE. + */ +int pvfs2_debugfs_init(void) +{ + + int rc = -ENOMEM; + + debug_dir = debugfs_create_dir("orangefs", NULL); + if (!debug_dir) + goto out; + + help_file_dentry = debugfs_create_file(ORANGEFS_KMOD_DEBUG_HELP_FILE, + 0444, + debug_dir, + debug_help_string, + &debug_help_fops); + if (!help_file_dentry) + goto out; + + orangefs_debug_disabled = 0; + rc = 0; + +out: + if (rc) + pvfs2_debugfs_cleanup(); + + return rc; +} + +void pvfs2_debugfs_cleanup(void) +{ + debugfs_remove_recursive(debug_dir); +} + +/* open ORANGEFS_KMOD_DEBUG_HELP_FILE */ +static int orangefs_debug_help_open(struct inode *inode, struct file *file) +{ + int rc = -ENODEV; + int ret; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_help_open: start\n"); + + if (orangefs_debug_disabled) + goto out; + + ret = seq_open(file, &help_debug_ops); + if (ret) + goto out; + + ((struct seq_file *)(file->private_data))->private = inode->i_private; + + rc = 0; + +out: + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_help_open: rc:%d:\n", + rc); + return rc; +} + +/* + * I think start always gets called again after stop. Start + * needs to return NULL when it is done. The whole "payload" + * in this case is a single (long) string, so by the second + * time we get to start (pos = 1), we're done. + */ +static void *help_start(struct seq_file *m, loff_t *pos) +{ + void *payload = NULL; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "help_start: start\n"); + + if (*pos == 0) + payload = m->private; + + return payload; +} + +static void *help_next(struct seq_file *m, void *v, loff_t *pos) +{ + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "help_next: start\n"); + + return NULL; +} + +static void help_stop(struct seq_file *m, void *p) +{ + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "help_stop: start\n"); +} + +static int help_show(struct seq_file *m, void *v) +{ + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "help_show: start\n"); + + seq_puts(m, v); + + return 0; +} + +/* + * initialize the kernel-debug file. + */ +int pvfs2_kernel_debug_init(void) +{ + + int rc = -ENOMEM; + struct dentry *ret; + char *k_buffer = NULL; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "%s: start\n", __func__); + + k_buffer = kzalloc(PVFS2_MAX_DEBUG_STRING_LEN, GFP_KERNEL); + if (!k_buffer) + goto out; + + if (strlen(kernel_debug_string) + 1 < PVFS2_MAX_DEBUG_STRING_LEN) { + strcpy(k_buffer, kernel_debug_string); + strcat(k_buffer, "\n"); + } else { + strcpy(k_buffer, "none\n"); + pr_info("%s: overflow 1!\n", __func__); + } + + ret = debugfs_create_file(ORANGEFS_KMOD_DEBUG_FILE, + 0444, + debug_dir, + k_buffer, + &kernel_debug_fops); + if (!ret) { + pr_info("%s: failed to create %s.\n", + __func__, + ORANGEFS_KMOD_DEBUG_FILE); + goto out; + } + + rc = 0; + +out: + if (rc) + pvfs2_debugfs_cleanup(); + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "%s: rc:%d:\n", __func__, rc); + return rc; +} + +/* + * initialize the client-debug file. + */ +int pvfs2_client_debug_init(void) +{ + + int rc = -ENOMEM; + char *c_buffer = NULL; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "%s: start\n", __func__); + + c_buffer = kzalloc(PVFS2_MAX_DEBUG_STRING_LEN, GFP_KERNEL); + if (!c_buffer) + goto out; + + if (strlen(client_debug_string) + 1 < PVFS2_MAX_DEBUG_STRING_LEN) { + strcpy(c_buffer, client_debug_string); + strcat(c_buffer, "\n"); + } else { + strcpy(c_buffer, "none\n"); + pr_info("%s: overflow! 2\n", __func__); + } + + client_debug_dentry = debugfs_create_file(ORANGEFS_CLIENT_DEBUG_FILE, + 0444, + debug_dir, + c_buffer, + &kernel_debug_fops); + if (!client_debug_dentry) { + pr_info("%s: failed to create %s.\n", + __func__, + ORANGEFS_CLIENT_DEBUG_FILE); + goto out; + } + + rc = 0; + +out: + if (rc) + pvfs2_debugfs_cleanup(); + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "%s: rc:%d:\n", __func__, rc); + return rc; +} + +/* open ORANGEFS_KMOD_DEBUG_FILE or ORANGEFS_CLIENT_DEBUG_FILE.*/ +int orangefs_debug_open(struct inode *inode, struct file *file) +{ + int rc = -ENODEV; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "%s: orangefs_debug_disabled: %d\n", + __func__, + orangefs_debug_disabled); + + if (orangefs_debug_disabled) + goto out; + + rc = 0; + mutex_lock(&orangefs_debug_lock); + file->private_data = inode->i_private; + mutex_unlock(&orangefs_debug_lock); + +out: + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_open: rc: %d\n", + rc); + return rc; +} + +static ssize_t orangefs_debug_read(struct file *file, + char __user *ubuf, + size_t count, + loff_t *ppos) +{ + char *buf; + int sprintf_ret; + ssize_t read_ret = -ENOMEM; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, "orangefs_debug_read: start\n"); + + buf = kmalloc(PVFS2_MAX_DEBUG_STRING_LEN, GFP_KERNEL); + if (!buf) + goto out; + + mutex_lock(&orangefs_debug_lock); + sprintf_ret = sprintf(buf, "%s", (char *)file->private_data); + mutex_unlock(&orangefs_debug_lock); + + read_ret = simple_read_from_buffer(ubuf, count, ppos, buf, sprintf_ret); + + kfree(buf); + +out: + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_read: ret: %zu\n", + read_ret); + + return read_ret; +} + +static ssize_t orangefs_debug_write(struct file *file, + const char __user *ubuf, + size_t count, + loff_t *ppos) +{ + char *buf; + int rc = -EFAULT; + size_t silly = 0; + char *debug_string; + struct pvfs2_kernel_op_s *new_op = NULL; + struct client_debug_mask c_mask = { NULL, 0, 0 }; + + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_write: %s\n", + file->f_path.dentry->d_name.name); + + /* + * Thwart users who try to jamb a ridiculous number + * of bytes into the debug file... + */ + if (count > PVFS2_MAX_DEBUG_STRING_LEN + 1) { + silly = count; + count = PVFS2_MAX_DEBUG_STRING_LEN + 1; + } + + buf = kmalloc(PVFS2_MAX_DEBUG_STRING_LEN, GFP_KERNEL); + if (!buf) + goto out; + memset(buf, 0, PVFS2_MAX_DEBUG_STRING_LEN); + + if (copy_from_user(buf, ubuf, count - 1)) { + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "%s: copy_from_user failed!\n", + __func__); + goto out; + } + + /* + * Map the keyword string from userspace into a valid debug mask. + * The mapping process involves mapping the human-inputted string + * into a valid mask, and then rebuilding the string from the + * verified valid mask. + * + * A service operation is required to set a new client-side + * debug mask. + */ + if (!strcmp(file->f_path.dentry->d_name.name, + ORANGEFS_KMOD_DEBUG_FILE)) { + debug_string_to_mask(buf, &gossip_debug_mask, 0); + debug_mask_to_string(&gossip_debug_mask, 0); + debug_string = kernel_debug_string; + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "New kernel debug string is %s\n", + kernel_debug_string); + } else { + /* Can't reset client debug mask if client is not running. */ + if (is_daemon_in_service()) { + pr_info("%s: Client not running :%d:\n", + __func__, + is_daemon_in_service()); + goto out; + } + + debug_string_to_mask(buf, &c_mask, 1); + debug_mask_to_string(&c_mask, 1); + debug_string = client_debug_string; + + new_op = op_alloc(PVFS2_VFS_OP_PARAM); + if (!new_op) { + pr_info("%s: op_alloc failed!\n", __func__); + goto out; + } + + new_op->upcall.req.param.op = + PVFS2_PARAM_REQUEST_OP_TWO_MASK_VALUES; + new_op->upcall.req.param.type = PVFS2_PARAM_REQUEST_SET; + memset(new_op->upcall.req.param.s_value, + 0, + PVFS2_MAX_DEBUG_STRING_LEN); + sprintf(new_op->upcall.req.param.s_value, + "%llx %llx\n", + c_mask.mask1, + c_mask.mask2); + + /* service_operation returns 0 on success... */ + rc = service_operation(new_op, + "pvfs2_param", + PVFS2_OP_INTERRUPTIBLE); + + if (rc) + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "%s: service_operation failed! rc:%d:\n", + __func__, + rc); + + op_release(new_op); + } + + mutex_lock(&orangefs_debug_lock); + memset(file->f_inode->i_private, 0, PVFS2_MAX_DEBUG_STRING_LEN); + sprintf((char *)file->f_inode->i_private, "%s\n", debug_string); + mutex_unlock(&orangefs_debug_lock); + + *ppos += count; + if (silly) + rc = silly; + else + rc = count; + +out: + gossip_debug(GOSSIP_DEBUGFS_DEBUG, + "orangefs_debug_write: rc: %d\n", + rc); + kfree(buf); + return rc; +} diff --git a/fs/orangefs/pvfs2-mod.c b/fs/orangefs/pvfs2-mod.c new file mode 100644 index 000000000000..9cbc992731d6 --- /dev/null +++ b/fs/orangefs/pvfs2-mod.c @@ -0,0 +1,316 @@ +/* + * (C) 2001 Clemson University and The University of Chicago + * + * Changes by Acxiom Corporation to add proc file handler for pvfs2 client + * parameters, Copyright Acxiom Corporation, 2005. + * + * See COPYING in top-level directory. + */ + +#include "protocol.h" +#include "pvfs2-kernel.h" +#include "pvfs2-debugfs.h" +#include "pvfs2-sysfs.h" + +/* PVFS2_VERSION is a ./configure define */ +#ifndef PVFS2_VERSION +#define PVFS2_VERSION "Unknown" +#endif + +/* + * global variables declared here + */ + +/* array of client debug keyword/mask values */ +struct client_debug_mask *cdm_array; +int cdm_element_count; + +char kernel_debug_string[PVFS2_MAX_DEBUG_STRING_LEN] = "none"; +char client_debug_string[PVFS2_MAX_DEBUG_STRING_LEN]; +char client_debug_array_string[PVFS2_MAX_DEBUG_STRING_LEN]; + +char *debug_help_string; +int help_string_initialized; +struct dentry *help_file_dentry; +struct dentry *client_debug_dentry; +struct dentry *debug_dir; +int client_verbose_index; +int client_all_index; +struct pvfs2_stats g_pvfs2_stats; + +/* the size of the hash tables for ops in progress */ +int hash_table_size = 509; + +static ulong module_parm_debug_mask; +__u64 gossip_debug_mask; +struct client_debug_mask client_debug_mask = { NULL, 0, 0 }; +unsigned int kernel_mask_set_mod_init; /* implicitly false */ +int op_timeout_secs = PVFS2_DEFAULT_OP_TIMEOUT_SECS; +int slot_timeout_secs = PVFS2_DEFAULT_SLOT_TIMEOUT_SECS; +__u32 DEBUG_LINE = 50; + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("PVFS2 Development Team"); +MODULE_DESCRIPTION("The Linux Kernel VFS interface to PVFS2"); +MODULE_PARM_DESC(module_parm_debug_mask, "debugging level (see pvfs2-debug.h for values)"); +MODULE_PARM_DESC(op_timeout_secs, "Operation timeout in seconds"); +MODULE_PARM_DESC(slot_timeout_secs, "Slot timeout in seconds"); +MODULE_PARM_DESC(hash_table_size, + "size of hash table for operations in progress"); + +static struct file_system_type pvfs2_fs_type = { + .name = "pvfs2", + .mount = pvfs2_mount, + .kill_sb = pvfs2_kill_sb, + .owner = THIS_MODULE, +}; + +module_param(hash_table_size, int, 0); +module_param(module_parm_debug_mask, ulong, 0755); +module_param(op_timeout_secs, int, 0); +module_param(slot_timeout_secs, int, 0); + +/* synchronizes the request device file */ +struct mutex devreq_mutex; + +/* + blocks non-priority requests from being queued for servicing. this + could be used for protecting the request list data structure, but + for now it's only being used to stall the op addition to the request + list +*/ +struct mutex request_mutex; + +/* hash table for storing operations waiting for matching downcall */ +struct list_head *htable_ops_in_progress; +DEFINE_SPINLOCK(htable_ops_in_progress_lock); + +/* list for queueing upcall operations */ +LIST_HEAD(pvfs2_request_list); + +/* used to protect the above pvfs2_request_list */ +DEFINE_SPINLOCK(pvfs2_request_list_lock); + +/* used for incoming request notification */ +DECLARE_WAIT_QUEUE_HEAD(pvfs2_request_list_waitq); + +static int __init pvfs2_init(void) +{ + int ret = -1; + __u32 i = 0; + + /* convert input debug mask to a 64-bit unsigned integer */ + gossip_debug_mask = (unsigned long long) module_parm_debug_mask; + + /* + * set the kernel's gossip debug string; invalid mask values will + * be ignored. + */ + debug_mask_to_string(&gossip_debug_mask, 0); + + /* remove any invalid values from the mask */ + debug_string_to_mask(kernel_debug_string, &gossip_debug_mask, 0); + + /* + * if the mask has a non-zero value, then indicate that the mask + * was set when the kernel module was loaded. The pvfs2 dev ioctl + * command will look at this boolean to determine if the kernel's + * debug mask should be overwritten when the client-core is started. + */ + if (gossip_debug_mask != 0) + kernel_mask_set_mod_init = true; + + /* print information message to the system log */ + pr_info("pvfs2: pvfs2_init called with debug mask: :%s: :%llx:\n", + kernel_debug_string, + (unsigned long long)gossip_debug_mask); + + ret = bdi_init(&pvfs2_backing_dev_info); + + if (ret) + return ret; + + if (op_timeout_secs < 0) + op_timeout_secs = 0; + + if (slot_timeout_secs < 0) + slot_timeout_secs = 0; + + /* initialize global book keeping data structures */ + ret = op_cache_initialize(); + if (ret < 0) + goto err; + + ret = dev_req_cache_initialize(); + if (ret < 0) + goto cleanup_op; + + ret = pvfs2_inode_cache_initialize(); + if (ret < 0) + goto cleanup_req; + + ret = kiocb_cache_initialize(); + if (ret < 0) + goto cleanup_inode; + + /* Initialize the pvfsdev subsystem. */ + ret = pvfs2_dev_init(); + if (ret < 0) { + gossip_err("pvfs2: could not initialize device subsystem %d!\n", + ret); + goto cleanup_kiocb; + } + + mutex_init(&devreq_mutex); + mutex_init(&request_mutex); + + htable_ops_in_progress = + kcalloc(hash_table_size, sizeof(struct list_head), GFP_KERNEL); + if (!htable_ops_in_progress) { + gossip_err("Failed to initialize op hashtable"); + ret = -ENOMEM; + goto cleanup_device; + } + + /* initialize a doubly linked at each hash table index */ + for (i = 0; i < hash_table_size; i++) + INIT_LIST_HEAD(&htable_ops_in_progress[i]); + + ret = fsid_key_table_initialize(); + if (ret < 0) + goto cleanup_progress_table; + + /* + * Build the contents of /sys/kernel/debug/orangefs/debug-help + * from the keywords in the kernel keyword/mask array. + * + * The keywords in the client keyword/mask array are + * unknown at boot time. + * + * orangefs_prepare_debugfs_help_string will be used again + * later to rebuild the debug-help file after the client starts + * and passes along the needed info. The argument signifies + * which time orangefs_prepare_debugfs_help_string is being + * called. + * + */ + ret = orangefs_prepare_debugfs_help_string(1); + if (ret) + goto out; + + pvfs2_debugfs_init(); + pvfs2_kernel_debug_init(); + orangefs_sysfs_init(); + + ret = register_filesystem(&pvfs2_fs_type); + if (ret == 0) { + pr_info("pvfs2: module version %s loaded\n", PVFS2_VERSION); + return 0; + } + + pvfs2_debugfs_cleanup(); + orangefs_sysfs_exit(); + fsid_key_table_finalize(); + +cleanup_progress_table: + kfree(htable_ops_in_progress); + +cleanup_device: + pvfs2_dev_cleanup(); + +cleanup_kiocb: + kiocb_cache_finalize(); + +cleanup_inode: + pvfs2_inode_cache_finalize(); + +cleanup_req: + dev_req_cache_finalize(); + +cleanup_op: + op_cache_finalize(); + +err: + bdi_destroy(&pvfs2_backing_dev_info); + +out: + return ret; +} + +static void __exit pvfs2_exit(void) +{ + int i = 0; + struct pvfs2_kernel_op_s *cur_op = NULL; + + gossip_debug(GOSSIP_INIT_DEBUG, "pvfs2: pvfs2_exit called\n"); + + unregister_filesystem(&pvfs2_fs_type); + pvfs2_debugfs_cleanup(); + orangefs_sysfs_exit(); + fsid_key_table_finalize(); + pvfs2_dev_cleanup(); + /* clear out all pending upcall op requests */ + spin_lock(&pvfs2_request_list_lock); + while (!list_empty(&pvfs2_request_list)) { + cur_op = list_entry(pvfs2_request_list.next, + struct pvfs2_kernel_op_s, + list); + list_del(&cur_op->list); + gossip_debug(GOSSIP_INIT_DEBUG, + "Freeing unhandled upcall request type %d\n", + cur_op->upcall.type); + op_release(cur_op); + } + spin_unlock(&pvfs2_request_list_lock); + + for (i = 0; i < hash_table_size; i++) + while (!list_empty(&htable_ops_in_progress[i])) { + cur_op = list_entry(htable_ops_in_progress[i].next, + struct pvfs2_kernel_op_s, + list); + op_release(cur_op); + } + + kiocb_cache_finalize(); + pvfs2_inode_cache_finalize(); + dev_req_cache_finalize(); + op_cache_finalize(); + + kfree(htable_ops_in_progress); + + bdi_destroy(&pvfs2_backing_dev_info); + + pr_info("pvfs2: module version %s unloaded\n", PVFS2_VERSION); +} + +/* + * What we do in this function is to walk the list of operations + * that are in progress in the hash table and mark them as purged as well. + */ +void purge_inprogress_ops(void) +{ + int i; + + for (i = 0; i < hash_table_size; i++) { + struct pvfs2_kernel_op_s *op; + struct pvfs2_kernel_op_s *next; + + list_for_each_entry_safe(op, + next, + &htable_ops_in_progress[i], + list) { + spin_lock(&op->lock); + gossip_debug(GOSSIP_INIT_DEBUG, + "pvfs2-client-core: purging in-progress op tag " + "%llu %s\n", + llu(op->tag), + get_opname_string(op)); + set_op_state_purged(op); + spin_unlock(&op->lock); + wake_up_interruptible(&op->waitq); + } + } +} + +module_init(pvfs2_init); +module_exit(pvfs2_exit);