*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2014, Intel Corporation.
+ * Copyright (c) 2011, 2016, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* Lustre is a trademark of Sun Microsystems, Inc.
*/
-#include <linux/kernel.h>
-#include <linux/mm.h>
-#include <linux/string.h>
-#include <linux/stat.h>
#include <linux/errno.h>
-#include <linux/unistd.h>
-#include <linux/version.h>
-#include <asm/uaccess.h>
-
-#include <linux/fs.h>
-#include <linux/stat.h>
-#include <asm/uaccess.h>
+#include <linux/delay.h>
+#include <linux/kernel.h>
#include <linux/mm.h>
-#include <linux/pagemap.h>
#define DEBUG_SUBSYSTEM S_LLITE
static const struct vm_operations_struct ll_file_vm_ops;
-void policy_from_vma(ldlm_policy_data_t *policy,
- struct vm_area_struct *vma, unsigned long addr,
- size_t count)
+void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
+ unsigned long addr, size_t count)
{
policy->l_extent.start = ((addr - vma->vm_start) & PAGE_MASK) +
- (vma->vm_pgoff << PAGE_CACHE_SHIFT);
- policy->l_extent.end = (policy->l_extent.start + count - 1) |
+ (vma->vm_pgoff << PAGE_SHIFT);
+ policy->l_extent.end = (policy->l_extent.start + count - 1) |
~PAGE_MASK;
}
/**
* API independent part for page fault initialization.
- * \param vma - virtual memory area addressed to page fault
* \param env - corespondent lu_env to processing
- * \param nest - nested level
+ * \param vma - virtual memory area addressed to page fault
* \param index - page index corespondent to fault.
* \parm ra_flags - vma readahead flags.
*
- * \return allocated and initialized env for fault operation.
- * \retval EINVAL if env can't allocated
- * \return other error codes from cl_io_init.
+ * \return error codes from cl_io_init.
*/
static struct cl_io *
-ll_fault_io_init(struct vm_area_struct *vma, struct lu_env **env_ret,
- struct cl_env_nest *nest, pgoff_t index,
- unsigned long *ra_flags)
+ll_fault_io_init(struct lu_env *env, struct vm_area_struct *vma,
+ pgoff_t index, unsigned long *ra_flags)
{
struct file *file = vma->vm_file;
- struct inode *inode = file->f_path.dentry->d_inode;
+ struct inode *inode = file_inode(file);
struct cl_io *io;
struct cl_fault_io *fio;
- struct lu_env *env;
int rc;
ENTRY;
- *env_ret = NULL;
if (ll_file_nolock(file))
RETURN(ERR_PTR(-EOPNOTSUPP));
- /*
- * page fault can be called when lustre IO is
- * already active for the current thread, e.g., when doing read/write
- * against user level buffer mapped from Lustre buffer. To avoid
- * stomping on existing context, optionally force an allocation of a new
- * one.
- */
- env = cl_env_nested_get(nest);
- if (IS_ERR(env))
- RETURN(ERR_PTR(-EINVAL));
-
- *env_ret = env;
-
+restart:
io = vvp_env_thread_io(env);
io->ci_obj = ll_i2info(inode)->lli_clob;
LASSERT(io->ci_obj != NULL);
} else {
LASSERT(rc < 0);
cl_io_fini(env, io);
- cl_env_nested_put(nest, env);
+ if (io->ci_need_restart)
+ goto restart;
+
io = ERR_PTR(rc);
}
- return io;
+ RETURN(io);
}
/* Sharing code of page_mkwrite method for rhel5 and rhel6 */
struct lu_env *env;
struct cl_io *io;
struct vvp_io *vio;
- struct cl_env_nest nest;
int result;
+ __u16 refcheck;
sigset_t set;
- struct inode *inode;
+ struct inode *inode = NULL;
struct ll_inode_info *lli;
ENTRY;
LASSERT(vmpage != NULL);
+ env = cl_env_get(&refcheck);
+ if (IS_ERR(env))
+ RETURN(PTR_ERR(env));
- io = ll_fault_io_init(vma, &env, &nest, vmpage->index, NULL);
+ io = ll_fault_io_init(env, vma, vmpage->index, NULL);
if (IS_ERR(io))
GOTO(out, result = PTR_ERR(io));
set = cfs_block_sigsinv(sigmask(SIGKILL) | sigmask(SIGTERM));
- /* we grab lli_trunc_sem to exclude truncate case.
- * Otherwise, we could add dirty pages into osc cache
- * while truncate is on-going. */
inode = vvp_object_inode(io->ci_obj);
lli = ll_i2info(inode);
- down_read(&lli->lli_trunc_sem);
result = cl_io_loop(env, io);
- up_read(&lli->lli_trunc_sem);
-
cfs_restore_sigs(set);
if (result == 0) {
out_io:
cl_io_fini(env, io);
- cl_env_nested_put(&nest, env);
out:
+ cl_env_put(env, &refcheck);
CDEBUG(D_MMAP, "%s mkwrite with %d\n", current->comm, result);
LASSERT(ergo(result == 0, PageLocked(vmpage)));
+ /* if page has been unmapped, presumably due to lock reclaim for
+ * concurrent usage, add some delay before retrying to prevent
+ * entering live-lock situation with competitors
+ */
+ if (result == -ENODATA && inode != NULL) {
+ CDEBUG(D_MMAP, "delaying new page-fault for inode %p to "
+ "prevent live-lock\n", inode);
+ msleep(10);
+ }
+
return result;
}
case 0:
result = VM_FAULT_LOCKED;
break;
- case -EFAULT:
- result = VM_FAULT_NOPAGE;
- break;
case -ENOMEM:
result = VM_FAULT_OOM;
break;
* \retval VM_FAULT_ERROR on general error
* \retval NOPAGE_OOM not have memory for allocate new page
*/
-static int ll_fault0(struct vm_area_struct *vma, struct vm_fault *vmf)
+static vm_fault_t ll_fault0(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- struct lu_env *env;
- struct cl_io *io;
- struct vvp_io *vio = NULL;
- struct page *vmpage;
- unsigned long ra_flags;
- struct cl_env_nest nest;
- int result;
- int fault_ret = 0;
- ENTRY;
+ struct lu_env *env;
+ struct cl_io *io;
+ struct vvp_io *vio = NULL;
+ struct page *vmpage;
+ unsigned long ra_flags;
+ int result = 0;
+ int fault_ret = 0;
+ __u16 refcheck;
+ ENTRY;
- io = ll_fault_io_init(vma, &env, &nest, vmf->pgoff, &ra_flags);
- if (IS_ERR(io))
- RETURN(to_fault_error(PTR_ERR(io)));
+ env = cl_env_get(&refcheck);
+ if (IS_ERR(env))
+ RETURN(PTR_ERR(env));
- result = io->ci_result;
+ if (ll_sbi_has_fast_read(ll_i2sbi(file_inode(vma->vm_file)))) {
+ /* do fast fault */
+ ll_cl_add(vma->vm_file, env, NULL, LCC_MMAP);
+ fault_ret = ll_filemap_fault(vma, vmf);
+ ll_cl_remove(vma->vm_file, env);
+
+ /* - If there is no error, then the page was found in cache and
+ * uptodate;
+ * - If VM_FAULT_RETRY is set, the page existed but failed to
+ * lock. It will return to kernel and retry;
+ * - Otherwise, it should try normal fault under DLM lock. */
+ if ((fault_ret & VM_FAULT_RETRY) ||
+ !(fault_ret & VM_FAULT_ERROR))
+ GOTO(out, result = 0);
+
+ fault_ret = 0;
+ }
+
+ io = ll_fault_io_init(env, vma, vmf->pgoff, &ra_flags);
+ if (IS_ERR(io))
+ GOTO(out, result = PTR_ERR(io));
+
+ result = io->ci_result;
if (result == 0) {
vio = vvp_env_io(env);
vio->u.fault.ft_vma = vma;
vio->u.fault.ft_flags_valid = 0;
/* May call ll_readpage() */
- ll_cl_add(vma->vm_file, env, io);
+ ll_cl_add(vma->vm_file, env, io, LCC_MMAP);
result = cl_io_loop(env, io);
vmpage = vio->u.fault.ft_vmpage;
if (result != 0 && vmpage != NULL) {
- page_cache_release(vmpage);
+ put_page(vmpage);
vmf->page = NULL;
}
}
cl_io_fini(env, io);
- cl_env_nested_put(&nest, env);
vma->vm_flags |= ra_flags;
+
+out:
+ cl_env_put(env, &refcheck);
if (result != 0 && !(fault_ret & VM_FAULT_RETRY))
fault_ret |= to_fault_error(result);
- CDEBUG(D_MMAP, "%s fault %d/%d\n",
- current->comm, fault_ret, result);
+ CDEBUG(D_MMAP, "%s fault %d/%d\n", current->comm, fault_ret, result);
RETURN(fault_ret);
}
-static int ll_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+#ifdef HAVE_VM_OPS_USE_VM_FAULT_ONLY
+static vm_fault_t ll_fault(struct vm_fault *vmf)
+{
+ struct vm_area_struct *vma = vmf->vma;
+#else
+static vm_fault_t ll_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
+#endif
int count = 0;
bool printed = false;
- int result;
+ bool cached;
+ vm_fault_t result;
sigset_t set;
+ ll_stats_ops_tally(ll_i2sbi(file_inode(vma->vm_file)),
+ LPROC_LL_FAULT, 1);
+
+ result = pcc_fault(vma, vmf, &cached);
+ if (cached)
+ return result;
+
/* Only SIGKILL and SIGTERM is allowed for fault/nopage/mkwrite
* so that it can be killed by admin but not cause segfault by
* other signals. */
set = cfs_block_sigsinv(sigmask(SIGKILL) | sigmask(SIGTERM));
+ /* make sure offset is not a negative number */
+ if (vmf->pgoff > (MAX_LFS_FILESIZE >> PAGE_SHIFT))
+ return VM_FAULT_SIGBUS;
restart:
- result = ll_fault0(vma, vmf);
- LASSERT(!(result & VM_FAULT_LOCKED));
- if (result == 0) {
+ result = ll_fault0(vma, vmf);
+ if (vmf->page &&
+ !(result & (VM_FAULT_RETRY | VM_FAULT_ERROR | VM_FAULT_LOCKED))) {
struct page *vmpage = vmf->page;
/* check if this page has been truncated */
lock_page(vmpage);
if (unlikely(vmpage->mapping == NULL)) { /* unlucky */
unlock_page(vmpage);
- page_cache_release(vmpage);
+ put_page(vmpage);
vmf->page = NULL;
if (!printed && ++count > 16) {
return result;
}
-static int ll_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
+#ifdef HAVE_VM_OPS_USE_VM_FAULT_ONLY
+static vm_fault_t ll_page_mkwrite(struct vm_fault *vmf)
{
- int count = 0;
- bool printed = false;
- bool retry;
- int result;
+ struct vm_area_struct *vma = vmf->vma;
+#else
+static vm_fault_t ll_page_mkwrite(struct vm_area_struct *vma,
+ struct vm_fault *vmf)
+{
+#endif
+ int count = 0;
+ bool printed = false;
+ bool retry;
+ bool cached;
+ vm_fault_t result;
+
+ ll_stats_ops_tally(ll_i2sbi(file_inode(vma->vm_file)),
+ LPROC_LL_MKWRITE, 1);
+
+ result = pcc_page_mkwrite(vma, vmf, &cached);
+ if (cached)
+ return result;
+ file_update_time(vma->vm_file);
do {
retry = false;
result = ll_page_mkwrite0(vma, vmf->page, &retry);
if (!printed && ++count > 16) {
- const struct dentry *de = vma->vm_file->f_path.dentry;
+ const struct dentry *de = file_dentry(vma->vm_file);
CWARN("app(%s): the page %lu of file "DFID" is under"
" heavy contention\n",
*/
static void ll_vm_open(struct vm_area_struct * vma)
{
- struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+ struct inode *inode = file_inode(vma->vm_file);
struct vvp_object *vob = cl_inode2vvp(inode);
ENTRY;
- LASSERT(vma->vm_file);
LASSERT(atomic_read(&vob->vob_mmap_cnt) >= 0);
atomic_inc(&vob->vob_mmap_cnt);
+ pcc_vm_open(vma);
EXIT;
}
*/
static void ll_vm_close(struct vm_area_struct *vma)
{
- struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+ struct inode *inode = file_inode(vma->vm_file);
struct vvp_object *vob = cl_inode2vvp(inode);
ENTRY;
- LASSERT(vma->vm_file);
atomic_dec(&vob->vob_mmap_cnt);
LASSERT(atomic_read(&vob->vob_mmap_cnt) >= 0);
+ pcc_vm_close(vma);
EXIT;
}
int rc = -ENOENT;
ENTRY;
- LASSERTF(last > first, "last "LPU64" first "LPU64"\n", last, first);
+ LASSERTF(last > first, "last %llu first %llu\n", last, first);
if (mapping_mapped(mapping)) {
rc = 0;
- unmap_mapping_range(mapping, first + PAGE_CACHE_SIZE - 1,
- last - first + 1, 0);
+ unmap_mapping_range(mapping, first + PAGE_SIZE - 1,
+ last - first + 1, 1);
}
RETURN(rc);
int ll_file_mmap(struct file *file, struct vm_area_struct * vma)
{
- struct inode *inode = file->f_path.dentry->d_inode;
+ struct inode *inode = file_inode(file);
+ bool cached;
int rc;
+
ENTRY;
if (ll_file_nolock(file))
RETURN(-EOPNOTSUPP);
+ rc = pcc_file_mmap(file, vma, &cached);
+ if (cached && rc != 0)
+ RETURN(rc);
+
ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_MAP, 1);
rc = generic_file_mmap(file, vma);
if (rc == 0) {
- vma->vm_ops = &ll_file_vm_ops;
+ vma->vm_ops = &ll_file_vm_ops;
vma->vm_ops->open(vma);
/* update the inode's size and mtime */
- rc = ll_glimpse_size(inode);
+ if (!cached)
+ rc = ll_glimpse_size(inode);
}
RETURN(rc);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff