static int ll_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
int count = 0;
bool printed = false;
int result;
sigset_t set;
/* 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));
ll_stats_ops_tally(ll_i2sbi(file_inode(vma->vm_file)),
LPROC_LL_FAULT, 1);
restart:
result = ll_fault0(vma, vmf);
if (!(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);
put_page(vmpage);
vmf->page = NULL;
if (!printed && ++count > 16) {
CWARN("the page is under heavy contention,"
"maybe your app(%s) needs revising :-)\n",
current->comm);
printed = true;
}
goto restart;
}
result |= VM_FAULT_LOCKED;
}
cfs_restore_sigs(set);
return result;
}
static int ll_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
int count = 0;
bool printed = false;
int result;
sigset_t set;
/* Only SIGKILL and SIGTERM are 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));
restart:
result = ll_fault0(vma, vmf);
LASSERT(!(result & VM_FAULT_LOCKED));
if (result == 0) {
struct page *vmpage = vmf->page;
/* check if this page has been truncated */
lock_page(vmpage);
if (unlikely(!vmpage->mapping)) { /* unlucky */
unlock_page(vmpage);
page_cache_release(vmpage);
vmf->page = NULL;
if (!printed && ++count > 16) {
CWARN("the page is under heavy contention, maybe your app(%s) needs revising :-)\n",
current->comm);
printed = true;
}
goto restart;
}
result = VM_FAULT_LOCKED;
}
cfs_restore_sigs(set);
return result;
}
/* Sharing code of page_mkwrite method for rhel5 and rhel6 */
static int ll_page_mkwrite0(struct vm_area_struct *vma, struct page *vmpage,
bool *retry)
{
struct lu_env *env;
struct cl_io *io;
struct vvp_io *vio;
struct cl_env_nest nest;
int result;
sigset_t set;
struct inode *inode;
struct ll_inode_info *lli;
io = ll_fault_io_init(vma, &env, &nest, vmpage->index, NULL);
if (IS_ERR(io)) {
result = PTR_ERR(io);
goto out;
}
result = io->ci_result;
if (result < 0)
goto out_io;
io->u.ci_fault.ft_mkwrite = 1;
io->u.ci_fault.ft_writable = 1;
vio = vvp_env_io(env);
vio->u.fault.ft_vma = vma;
vio->u.fault.ft_vmpage = vmpage;
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 = ccc_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) {
struct inode *inode = file_inode(vma->vm_file);
struct ll_inode_info *lli = ll_i2info(inode);
lock_page(vmpage);
if (!vmpage->mapping) {
unlock_page(vmpage);
/* page was truncated and lock was cancelled, return
* ENODATA so that VM_FAULT_NOPAGE will be returned
* to handle_mm_fault().
*/
if (result == 0)
result = -ENODATA;
} else if (!PageDirty(vmpage)) {
/* race, the page has been cleaned by ptlrpcd after
* it was unlocked, it has to be added into dirty
* cache again otherwise this soon-to-dirty page won't
* consume any grants, even worse if this page is being
* transferred because it will break RPC checksum.
*/
unlock_page(vmpage);
CDEBUG(D_MMAP, "Race on page_mkwrite %p/%lu, page has been written out, retry.\n",
vmpage, vmpage->index);
*retry = true;
result = -EAGAIN;
}
if (result == 0) {
spin_lock(&lli->lli_lock);
lli->lli_flags |= LLIF_DATA_MODIFIED;
spin_unlock(&lli->lli_lock);
}
}
out_io:
cl_io_fini(env, io);
cl_env_nested_put(&nest, env);
out:
CDEBUG(D_MMAP, "%s mkwrite with %d\n", current->comm, result);
LASSERT(ergo(result == 0, PageLocked(vmpage)));
return result;
}
/* Sharing code of page_mkwrite method for rhel5 and rhel6 */
static int ll_page_mkwrite0(struct vm_area_struct *vma, struct page *vmpage,
bool *retry)
{
struct lu_env *env;
struct cl_io *io;
struct vvp_io *vio;
int result;
__u16 refcheck;
sigset_t set;
struct inode *inode;
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(env, vma, vmpage->index, NULL);
if (IS_ERR(io))
GOTO(out, result = PTR_ERR(io));
result = io->ci_result;
if (result < 0)
GOTO(out_io, result);
io->u.ci_fault.ft_mkwrite = 1;
io->u.ci_fault.ft_writable = 1;
vio = vvp_env_io(env);
vio->u.fault.ft_vma = vma;
vio->u.fault.ft_vmpage = vmpage;
set = cfs_block_sigsinv(sigmask(SIGKILL) | sigmask(SIGTERM));
inode = vvp_object_inode(io->ci_obj);
lli = ll_i2info(inode);
result = cl_io_loop(env, io);
cfs_restore_sigs(set);
if (result == 0) {
lock_page(vmpage);
if (vmpage->mapping == NULL) {
unlock_page(vmpage);
/* page was truncated and lock was cancelled, return
* ENODATA so that VM_FAULT_NOPAGE will be returned
* to handle_mm_fault(). */
if (result == 0)
result = -ENODATA;
} else if (!PageDirty(vmpage)) {
/* race, the page has been cleaned by ptlrpcd after
* it was unlocked, it has to be added into dirty
* cache again otherwise this soon-to-dirty page won't
* consume any grants, even worse if this page is being
* transferred because it will break RPC checksum.
*/
unlock_page(vmpage);
CDEBUG(D_MMAP, "Race on page_mkwrite %p/%lu, page has "
"been written out, retry.\n",
vmpage, vmpage->index);
*retry = true;
result = -EAGAIN;
}
if (result == 0)
ll_file_set_flag(lli, LLIF_DATA_MODIFIED);
}
EXIT;
out_io:
cl_io_fini(env, io);
out:
cl_env_put(env, &refcheck);
CDEBUG(D_MMAP, "%s mkwrite with %d\n", current->comm, result);
LASSERT(ergo(result == 0, PageLocked(vmpage)));
return result;
}
