maildir-sync.c revision 406a92a344f789e6ec763104b03de3f0b8fcfdb7
/* Copyright (C) 2002-2003 Timo Sirainen */
/*
Here's a description of how we handle Maildir synchronization and
it's problems:
We want to be as efficient as we can. The most efficient way to
check if changes have occured is to stat() the new/ and cur/
directories and uidlist file - if their mtimes haven't changed,
there's no changes and we don't need to do anything.
Problem 1: Multiple changes can happen within a single second -
nothing guarantees that once we synced it, someone else didn't just
then make a modification. Such modifications wouldn't get noticed
until a new modification occured later.
Problem 2: Syncing cur/ directory is much more costly than syncing
new/. Moving mails from new/ to cur/ will always change mtime of
cur/ causing us to sync it as well.
Problem 3: We may not be able to move mail from new/ to cur/
because we're out of quota, or simply because we're accessing a
read-only mailbox.
MAILDIR_SYNC_SECS
-----------------
Several checks below use MAILDIR_SYNC_SECS, which should be maximum
clock drift between all computers accessing the maildir (eg. via
NFS), rounded up to next second. Our default is 1 second, since
everyone should be using NTP.
Note that setting it to 0 works only if there's only one computer
accessing the maildir. It's practically impossible to make two
clocks _exactly_ synchronized.
It might be possible to only use file server's clock by looking at
the atime field, but I don't know how well that would actually work.
cur directory
-------------
We have maildir_cur_dirty variable which is set to cur/ directory's
mtime when it's >= time() - MAILDIR_SYNC_SECS and we _think_ we have
synchronized the directory.
When maildir_cur_dirty is non-zero, we don't synchronize the cur/
directory until
a) cur/'s mtime changes
b) opening a mail fails with ENOENT
c) time() > maildir_cur_dirty + MAILDIR_SYNC_SECS
This allows us to modify the maildir multiple times without having
to sync it at every change. The sync will eventually be done to
make sure we didn't miss any external changes.
The maildir_cur_dirty is set when:
- we change message flags
- we expunge messages
- we move mail from new/ to cur/
- we sync cur/ directory and it's mtime is
>= time() - MAILDIR_SYNC_SECS
It's unset when we do the final syncing, ie. when mtime is
older than time() - MAILDIR_SYNC_SECS.
new directory
-------------
If new/'s mtime is >= time() - MAILDIR_SYNC_SECS, always synchronize
it. maildir_cur_dirty-like feature might save us a few syncs, but
that might break a client which saves a mail in one connection and
tries to fetch it in another one. new/ directory is almost always
empty, so syncing it should be very fast anyway. Actually this can
still happen if we sync only new/ dir while another client is also
moving mails from it to cur/ - it takes us a while to see them.
That's pretty unlikely to happen however, and only way to fix it
would be to always synchronize cur/ after new/.
Normally we move all mails from new/ to cur/ whenever we sync it. If
it's not possible for some reason, we set maildir_have_new flag on
which instructs synchronization to check files in new/ directory as
well. maildir_keep_new flag is also set which instructs syncing to
not even try to move mails to cur/ anymore.
If client tries to change a flag for message in new/, we try to
rename() it into cur/. If it's successful, we clear the
maildir_keep_new flag so at next sync we'll try to move all of them
to cur/. When all of them have been moved, maildir_have_new flag is
cleared as well. Expunges will also clear maildir_keep_new flag.
If rename() still fails because of ENOSPC or EDQUOT, we still save
the flag changes in index with dirty-flag on. When moving the mail
to cur/ directory, or when we notice it's already moved there, we
apply the flag changes to the filename, rename it and remove the
dirty flag. If there's dirty flags, this should be tried every time
after expunge or when closing the mailbox.
uidlist
-------
This file contains UID <-> filename mappings. It's updated only when
new mail arrives, so it may contain filenames that have already been
deleted. Updating is done by getting uidlist.lock file, writing the
whole uidlist into it and rename()ing it over the old uidlist. This
means there's no need to lock the file for reading.
Whenever uidlist is rewritten, it's mtime must be larger than the old
one's. Use utime() before rename() if needed.
Only time you have to read this file is when assigning new UIDs for
messages, to see if they already have UIDs. If file's mtime hasn't
changed, you don't have to do even that.
broken clients
--------------
Originally the middle identifier in Maildir filename was specified
only as <process id>_<delivery counter>. That however created a
problem with randomized PIDs which made it possible that the same
PID was reused within one second.
So if within one second a mail was delivered, MUA moved it to cur/
and another mail was delivered by a new process using same PID as
the first one, we likely ended up overwriting the first mail when
the second mail was moved over it.
Nowadays everyone should be giving a bit more specific identifier,
for example include microseconds in it which Dovecot does.
There's a simple way to prevent this from happening in some cases:
Don't move the mail from new/ to cur/ if it's mtime is >= time() -
MAILDIR_SYNC_SECS. The second delivery's link() call then fails
because the file is already in new/, and it will then use a
different filename. There's a few problems with this however:
- it requires extra stat() call which is unneeded extra I/O
- another MUA might still move the mail to cur/
- if first file's flags are modified by either Dovecot or another
MUA, it's moved to cur/ (you _could_ just do the dirty-flagging
but that'd be ugly)
Because this is useful only for very few people and it requires
extra I/O, I decided not to implement this. It should be however
quite easy to do since we need to be able to deal with files in new/
in any case.
It's also possible to never accidentally overwrite a mail by using
link() + unlink() rather than rename(). This however isn't very
good idea as it introduces potential race conditions when multiple
clients are accessing the mailbox:
Trying to move the same mail from new/ to cur/ at the same time:
a) Client 1 uses slightly different filename than client 2,
for example one sets read-flag on but the other doesn't.
You have the same mail duplicated now.
b) Client 3 sees the mail between Client 1's and 2's link() calls
and changes it's flag. You have the same mail duplicated now.
And it gets worse when they're unlink()ing in cur/ directory:
c) Client 1 changes mails's flag and client 2 changes it back
between 1's link() and unlink(). The mail is now expunged.
d) If you try to deal with the duplicates by unlink()ing another
one of them, you might end up unlinking both of them.
So, what should we do then if we notice a duplicate? First of all,
it might not be a duplicate at all, readdir() might have just
returned it twice because it was just renamed. What we should do is
create a completely new base name for it and rename() it to that.
If the call fails with ENOENT, it only means that it wasn't a
duplicate after all.
*/
#include "lib.h"
#include "buffer.h"
#include "istream.h"
#include "hash.h"
#include "ioloop.h"
#include "str.h"
#include "maildir-index.h"
#include "maildir-uidlist.h"
#include "mail-index-data.h"
#include "mail-index-util.h"
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <dirent.h>
#include <utime.h>
#define MAILDIR_SYNC_SECS 1
enum maildir_file_action {
MAILDIR_FILE_FLAG_NEWDIR = 0x1000,
MAILDIR_FILE_FLAG_ALLOCED = 0x2000,
MAILDIR_FILE_FLAGS = 0x3000
};
struct maildir_hash_context {
struct mail_index *index;
struct mail_index_record *new_mail;
int failed;
};
struct maildir_hash_rec {
struct mail_index_record *rec;
enum maildir_file_action action;
};
struct maildir_sync_context {
struct mail_index *index;
struct hash_table *files;
unsigned int new_count;
struct maildir_uidlist *uidlist;
unsigned int readonly_check:1;
unsigned int flag_updates:1;
unsigned int uidlist_rewrite:1;
unsigned int new_mails_new:1;
unsigned int new_mails_cur:1;
};
/* a char* hash function from ASU -- from glib */
static unsigned int maildir_hash(const void *p)
{
const unsigned char *s = p;
unsigned int g, h = 0;
while (*s != ':' && *s != '\0') {
h = (h << 4) + *s;
if ((g = h & 0xf0000000UL)) {
h = h ^ (g >> 24);
h = h ^ g;
}
s++;
}
return h;
}
{
}
return 0;
}
const char *fname)
{
char *new_fname;
}
}
}
struct mail_index_record *rec,
const char *new_fname)
{
struct mail_index_update *update;
return TRUE;
}
}
struct mail_index_record *rec,
{
enum mail_flags flags;
return TRUE;
return FALSE;
}
return TRUE;
}
{
const char *path;
return TRUE;
/* open it only if it's changed since we last synced it. */
/* doesn't exist yet, create it */
case -1:
return FALSE;
case 1:
break;
}
return TRUE;
}
}
/* FIXME: last_uidlist_mtime should be in index headers */
return TRUE;
return TRUE;
/* uidvalidity changed */
"UIDVALIDITY changed in uidlist");
return FALSE;
}
if (!index->rebuilding) {
return FALSE;
}
}
"uidlist.next_uid (%u)",
return FALSE;
}
return TRUE;
}
struct mail_index_record *rec,
{
struct mail_index_data_record_header *data_hdr;
const char *path;
/* nothing in cache, we can't know if it's changed */
return FALSE;
}
t_push();
t_pop();
return FALSE;
}
t_pop();
return FALSE;
return TRUE;
return TRUE;
return FALSE;
}
{
}
{
struct maildir_uidlist *uidlist;
struct mail_index_record *rec;
struct maildir_hash_rec *hash_rec;
struct maildir_uidlist_rec uid_rec;
enum maildir_file_action action;
void *orig_key, *orig_value;
int new_dir;
/* new mails, either they're already in uidlist or we have
to add them there. If we want to add them, we'll need to
sync it locked. */
return FALSE;
if (!maildir_sync_open_uidlist(ctx))
return FALSE;
}
seq = 0;
else {
return FALSE;
}
/* skip over the expunged records in uidlist */
return FALSE;
}
return FALSE;
&orig_key, &orig_value)) {
/* none action */
} else {
}
"Filename mismatch for UID %u: %s vs %s",
return FALSE;
}
/* it's UID has changed. shouldn't happen. */
i_warning("UID changed for %s/%s: %u -> %u",
}
switch (action) {
return FALSE;
seq--;
break;
return FALSE;
return FALSE;
break;
return FALSE;
seq--;
break;
case MAILDIR_FILE_ACTION_NONE:
break;
default:
i_unreached();
}
return FALSE;
}
}
"(%u != %u)", seq,
return FALSE;
}
/* if there's new mails which are already in uidlist, get them */
last_uid = 0;
&orig_key, &orig_value))
else
/* it's a duplicate, shouldn't happen */
i_error("%s: Found duplicate filename %s, rebuilding",
if (INDEX_IS_UIDLIST_LOCKED(index))
}
/* make sure we set the same UID for it. */
"index.next_uid (%u) > "
"uid_rec.uid (%u)",
return FALSE;
}
if (new_flag != 0)
new_flag != 0))
return FALSE;
}
return FALSE;
}
/* update our next_uid. it should have been checked for
sanity already. */
/* uidlist is now synced, remember that. */
return index_file_set_syscall_error(index,
"fstat()");
}
}
/* all done (or can't do it since we don't have lock) */
return TRUE;
}
/* then there's the completely new mails. sort them by the filename
so we should get them to same order as they were created. */
(int (*)(const void *, const void *)) strcmp);
if (!index->maildir_keep_new) {
} else {
/* this is actually slightly wrong, because we don't really
know if some of the new messages are in cur/ already.
we could know that by saving it into buffer, but that'd
require extra memory. luckily it doesn't really matter if
we say it's in new/, but it's actually in cur/. we have
to deal with such case anyway since another client might
have just moved it. */
}
return FALSE;
}
return TRUE;
}
int only_new)
{
struct mail_index_record *rec;
struct maildir_hash_rec *hash_rec;
const char *fname;
int have_new;
/* FIXME: kludge. we want to have pointers to data file, so we must
make sure that it's base address doesn't change. this call makes
sure it's fully mmaped in memory even when we begin */
return FALSE;
return FALSE;
}
/* read current messages in index into hash */
sizeof(struct maildir_hash_rec) + 1024);
return FALSE;
if (!only_new ||
"Duplicated message %s", fname);
return FALSE;
}
/* WARNING: index must not be modified as long as
these hash keys exist. Modifying might change the
mmap base address. */
}
}
return TRUE;
}
{
t_push();
i_warning("Fixed duplicate in %s: %s -> %s",
}
t_pop();
return ret;
}
{
case MAILDIR_FILE_ACTION_NONE:
break;
/* we're getting here because our recently
inserted node is traversed as well */
break;
}
break;
default:
break;
}
}
{
struct maildir_hash_rec *hash_rec;
void *orig_key, *orig_value;
int check_content_changes, newflag;
/* Do we want to check changes in file contents? This slows down
things as we need to do extra stat() for all files. */
do {
if (d->d_name[0] == '.')
continue;
&orig_key, &orig_value)) {
} else {
return FALSE;
continue;
}
}
/* new message */
if (ctx->readonly_check)
continue;
if (new_dir)
else
hash_rec);
continue;
}
INDEX_MAIL_FLAG_MAILDIR_NEW) != 0 &&
/* mail was indexed in new/ but it has been
moved to cur/ later */
}
if (check_content_changes &&
/* file content changed, treat it as new message */
hash_rec);
hash_rec);
} else {
}
/* records that are left to hash must not have any (filename) pointers
to index file. So remove none actions, and p_strdup() expunge
actions. */
return TRUE;
}
{
struct dirent *d;
"opendir()");
}
/* find first file */
if (d->d_name[0] != '.')
break;
}
if (d == NULL) {
"closedir()");
}
} else {
}
return TRUE;
}
{
int failed;
if (!maildir_new_scan_first_file(ctx))
return FALSE;
}
return FALSE;
}
"opendir()");
}
"closedir()");
}
return !failed;
}
{
return FALSE;
return FALSE;
return FALSE;
if (!maildir_full_sync_finish(ctx))
return FALSE;
return TRUE;
}
int move_to_cur, int append_index)
{
struct dirent *d;
const char *final_dir;
if (append_index) {
/* some of the mails in new/ are already indexed.
we'll have to do a full sync. */
return maildir_sync_new_dir_full(ctx);
}
return FALSE;
case -1:
return FALSE;
case 0:
/* couldn't get a lock.
no point in doing more. */
return TRUE;
}
/* make sure uidlist is up to date.
if it's not, do a full sync. */
if (!maildir_sync_open_uidlist(ctx))
return FALSE;
return maildir_sync_cur_dir(ctx);
}
do {
if (d->d_name[0] == '.')
continue;
str_truncate(sourcepath, 0);
if (move_to_cur) {
str_truncate(destpath, 0);
else {
"rename(%s, %s) failed: %m",
return FALSE;
}
if (!append_index) {
return TRUE;
}
/* continue by keeping them in new/ dir */
move_to_cur = FALSE;
}
}
if (append_index) {
if (!move_to_cur)
t_push();
d->d_name,
!move_to_cur)) {
t_pop();
return FALSE;
}
t_pop();
}
return TRUE;
}
{
/* there's also new mails. move them into cur/ first, if we
can lock the uidlist */
switch (maildir_uidlist_try_lock(index)) {
case -1:
return FALSE;
case 1:
return FALSE;
}
}
return FALSE;
return FALSE;
return TRUE;
}
int *changes)
{
return FALSE;
/* FIXME: file_sync_stamp should be in index file's headers.
it should also contain maildir_cur_dirty. */
}
return FALSE;
}
return FALSE;
}
if (!maildir_new_scan_first_file(ctx))
return FALSE;
}
(index->maildir_cur_dirty != 0 &&
/* cur/ changed, or delayed cur/ check */
if (!maildir_sync_cur_dir(ctx))
return FALSE;
}
return FALSE;
/* this will set maildir_cur_dirty. it may actually be
different from cur/'s mtime if we're unlucky, but that
doesn't really matter and it's not worth the extra stat() */
}
if (ctx->uidlist_rewrite) {
return FALSE;
}
if (index->maildir_have_new)
else
}
if (index->maildir_cur_dirty == 0 ||
index->maildir_cur_dirty = 0;
else {
/* uidlist is locked, wait for a while before
trying again */
}
}
return TRUE;
}
{
struct mail_index_record *rec;
struct maildir_hash_rec *hash_rec;
void *orig_key, *orig_value;
const char *fname;
unsigned int seq;
return TRUE;
return FALSE;
else
return FALSE;
return FALSE;
}
}
return TRUE;
}
{
int cur_changed;
return FALSE;
}
index->maildir_cur_dirty != 0;
if (!cur_changed) {
if (!index->maildir_have_new) {
/* no changes */
return TRUE;
}
"stat()");
}
/* no changes */
return TRUE;
}
if (!maildir_new_scan_first_file(ctx))
return FALSE;
}
/* ok, something's changed. check only changes in file names. */
/* if we can get exclusive lock, we can update the index
directly. but don't rely on it. */
return FALSE;
return TRUE;
}
{
"closedir()");
}
}
}
static struct maildir_sync_context *
{
struct maildir_sync_context *ctx;
}
return ctx;
}
{
struct maildir_sync_context *ctx;
struct maildir_hash_rec *hash_rec;
int ret;
if (!ret)
else {
}
return ret;
}
int *changes)
{
struct maildir_sync_context *ctx;
int ret;
return ret;
}