/**
* Lustre implementation of a vm_operations_struct::fault() method, called by
* VM to server page fault (both in kernel and user space).
*
* \param vma - is virtual area struct related to page fault
* \param vmf - structure which describe type and address where hit fault
*
* \return allocated and filled _locked_ page for address
* \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)
{
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;
io = ll_fault_io_init(vma, &env, &nest, vmf->pgoff, &ra_flags);
if (IS_ERR(io))
return to_fault_error(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_vmpage = NULL;
vio->u.fault.ft_vmf = vmf;
vio->u.fault.ft_flags = 0;
vio->u.fault.ft_flags_valid = false;
/* May call ll_readpage() */
ll_cl_add(vma->vm_file, env, io);
result = cl_io_loop(env, io);
ll_cl_remove(vma->vm_file, env);
/* ft_flags are only valid if we reached
* the call to filemap_fault
*/
if (vio->u.fault.ft_flags_valid)
fault_ret = vio->u.fault.ft_flags;
vmpage = vio->u.fault.ft_vmpage;
if (result != 0 && vmpage) {
put_page(vmpage);
vmf->page = NULL;
}
}
cl_io_fini(env, io);
cl_env_nested_put(&nest, env);
vma->vm_flags |= ra_flags;
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);
return fault_ret;
}
/* 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;
}
/**
* Lustre implementation of a vm_operations_struct::fault() method, called by
* VM to server page fault (both in kernel and user space).
*
* \param vma - is virtiual area struct related to page fault
* \param vmf - structure which describe type and address where hit fault
*
* \return allocated and filled _locked_ page for address
* \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)
{
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;
env = cl_env_get(&refcheck);
if (IS_ERR(env))
RETURN(PTR_ERR(env));
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 = 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_vmpage = NULL;
vio->u.fault.ft_vmf = vmf;
vio->u.fault.ft_flags = 0;
vio->u.fault.ft_flags_valid = 0;
/* May call ll_readpage() */
ll_cl_add(vma->vm_file, env, io, LCC_MMAP);
result = cl_io_loop(env, io);
ll_cl_remove(vma->vm_file, env);
/* ft_flags are only valid if we reached
* the call to filemap_fault */
if (vio->u.fault.ft_flags_valid)
fault_ret = vio->u.fault.ft_flags;
vmpage = vio->u.fault.ft_vmpage;
if (result != 0 && vmpage != NULL) {
put_page(vmpage);
vmf->page = NULL;
}
}
cl_io_fini(env, io);
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);
RETURN(fault_ret);
}
/* 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;
}