maildir-sync.c revision ead79af955bccc360de08c150918474c1809748e
3853N/A/* Copyright (C) 2004 Timo Sirainen */
3853N/A
3853N/A/*
3853N/A Here's a description of how we handle Maildir synchronization and
3853N/A it's problems:
3853N/A
3853N/A We want to be as efficient as we can. The most efficient way to
3853N/A check if changes have occured is to stat() the new/ and cur/
6983N/A directories and uidlist file - if their mtimes haven't changed,
6983N/A there's no changes and we don't need to do anything.
3853N/A
3853N/A Problem 1: Multiple changes can happen within a single second -
3853N/A nothing guarantees that once we synced it, someone else didn't just
3853N/A then make a modification. Such modifications wouldn't get noticed
6983N/A until a new modification occured later.
6983N/A
6983N/A Problem 2: Syncing cur/ directory is much more costly than syncing
6983N/A new/. Moving mails from new/ to cur/ will always change mtime of
3853N/A cur/ causing us to sync it as well.
3853N/A
3853N/A Problem 3: We may not be able to move mail from new/ to cur/
3853N/A because we're out of quota, or simply because we're accessing a
3853N/A read-only mailbox.
3853N/A
3853N/A
3853N/A MAILDIR_SYNC_SECS
3853N/A -----------------
3853N/A
3853N/A Several checks below use MAILDIR_SYNC_SECS, which should be maximum
3853N/A clock drift between all computers accessing the maildir (eg. via
3853N/A NFS), rounded up to next second. Our default is 1 second, since
3853N/A everyone should be using NTP.
3853N/A
3853N/A Note that setting it to 0 works only if there's only one computer
3853N/A accessing the maildir. It's practically impossible to make two
3853N/A clocks _exactly_ synchronized.
3853N/A
3853N/A It might be possible to only use file server's clock by looking at
3853N/A the atime field, but I don't know how well that would actually work.
3853N/A
3853N/A cur directory
3853N/A -------------
3853N/A
3853N/A We have dirty_cur_time variable which is set to cur/ directory's
3853N/A mtime when it's >= time() - MAILDIR_SYNC_SECS and we _think_ we have
3853N/A synchronized the directory.
3853N/A
3853N/A When dirty_cur_time is non-zero, we don't synchronize the cur/
3853N/A directory until
3853N/A
3853N/A a) cur/'s mtime changes
3853N/A b) opening a mail fails with ENOENT
3853N/A c) time() > dirty_cur_time + MAILDIR_SYNC_SECS
3853N/A
3853N/A This allows us to modify the maildir multiple times without having
3853N/A to sync it at every change. The sync will eventually be done to
3853N/A make sure we didn't miss any external changes.
3853N/A
3853N/A The dirty_cur_time is set when:
3853N/A
3853N/A - we change message flags
3853N/A - we expunge messages
3853N/A - we move mail from new/ to cur/
3853N/A - we sync cur/ directory and it's mtime is >= time() - MAILDIR_SYNC_SECS
3853N/A
3853N/A It's unset when we do the final syncing, ie. when mtime is
3853N/A older than time() - MAILDIR_SYNC_SECS.
3853N/A
3853N/A new directory
3853N/A -------------
3853N/A
3853N/A If new/'s mtime is >= time() - MAILDIR_SYNC_SECS, always synchronize
3853N/A it. dirty_cur_time-like feature might save us a few syncs, but
3853N/A that might break a client which saves a mail in one connection and
3853N/A tries to fetch it in another one. new/ directory is almost always
3853N/A 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 mark the mail with "probably
exists in new/ directory" 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. Note that inode checking
wouldn't have been sufficient as inode numbers can be reused.
This file is usually read the first time you need to know filename for
given UID. After that it's not re-read unless new mails come that we
don't know about.
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 "ioloop.h"
#include "buffer.h"
#include "hash.h"
#include "str.h"
#include "maildir-storage.h"
#include "maildir-uidlist.h"
#include <stdio.h>
#include <unistd.h>
#include <dirent.h>
#include <sys/stat.h>
#define MAILDIR_SYNC_SECS 1
#define MAILDIR_FILENAME_FLAG_FOUND 128
struct maildir_sync_context {
struct index_mailbox *ibox;
const char *new_dir, *cur_dir;
int partial;
struct maildir_uidlist_sync_ctx *uidlist_sync_ctx;
};
struct maildir_index_sync_context {
struct index_mailbox *ibox;
struct mail_index_view *view;
struct mail_index_sync_ctx *sync_ctx;
struct mail_index_sync_rec sync_rec;
uint32_t seq;
};
static int maildir_expunge(struct index_mailbox *ibox, const char *path,
void *context __attr_unused__)
{
if (unlink(path) == 0) {
ibox->dirty_cur_time = ioloop_time;
return 1;
}
if (errno == ENOENT)
return 0;
mail_storage_set_critical(ibox->box.storage,
"unlink(%s) failed: %m", path);
return -1;
}
static int maildir_sync_flags(struct index_mailbox *ibox, const char *path,
void *context)
{
struct maildir_index_sync_context *ctx = context;
const char *newpath;
enum mail_flags flags;
uint8_t flags8;
keywords_mask_t keywords;
(void)maildir_filename_get_flags(path, &flags, keywords);
flags8 = flags;
mail_index_sync_flags_apply(&ctx->sync_rec, &flags8, keywords);
newpath = maildir_filename_set_flags(path, flags8, keywords);
if (rename(path, newpath) == 0) {
ibox->dirty_cur_time = ioloop_time;
return 1;
}
if (errno == ENOENT)
return 0;
if (ENOSPACE(errno)) {
if (mail_index_sync_set_dirty(ctx->sync_ctx, ctx->seq) < 0)
return -1;
return 1;
}
mail_storage_set_critical(ibox->box.storage,
"rename(%s, %s) failed: %m", path, newpath);
return -1;
}
static int maildir_sync_record(struct index_mailbox *ibox,
struct maildir_index_sync_context *ctx)
{
struct mail_index_sync_rec *sync_rec = &ctx->sync_rec;
struct mail_index_view *view = ctx->view;
uint32_t seq, uid;
switch (sync_rec->type) {
case MAIL_INDEX_SYNC_TYPE_APPEND:
break;
case MAIL_INDEX_SYNC_TYPE_EXPUNGE:
for (seq = sync_rec->seq1; seq <= sync_rec->seq2; seq++) {
if (mail_index_lookup_uid(view, seq, &uid) < 0)
return -1;
if (maildir_file_do(ibox, uid, maildir_expunge,
NULL) < 0)
return -1;
}
break;
case MAIL_INDEX_SYNC_TYPE_FLAGS:
ctx->seq = sync_rec->seq1;
for (; ctx->seq <= sync_rec->seq2; ctx->seq++) {
if (mail_index_lookup_uid(view, ctx->seq, &uid) < 0)
return -1;
if (maildir_file_do(ibox, uid, maildir_sync_flags,
ctx) < 0)
return -1;
}
break;
}
return 0;
}
int maildir_sync_last_commit(struct index_mailbox *ibox)
{
struct maildir_index_sync_context ctx;
int ret;
if (ibox->commit_log_file_seq == 0)
return 0;
memset(&ctx, 0, sizeof(ctx));
ctx.ibox = ibox;
ret = mail_index_sync_begin(ibox->index, &ctx.sync_ctx, &ctx.view,
ibox->commit_log_file_seq,
ibox->commit_log_file_offset);
if (ret > 0) {
while ((ret = mail_index_sync_next(ctx.sync_ctx,
&ctx.sync_rec)) > 0) {
if (maildir_sync_record(ibox, &ctx) < 0) {
ret = -1;
break;
}
}
if (mail_index_sync_end(ctx.sync_ctx, 0, 0) < 0)
ret = -1;
}
if (ret == 0) {
ibox->commit_log_file_seq = 0;
ibox->commit_log_file_offset = 0;
} else {
mail_storage_set_index_error(ibox);
}
return ret;
}
static struct maildir_sync_context *
maildir_sync_context_new(struct index_mailbox *ibox)
{
struct maildir_sync_context *ctx;
ctx = t_new(struct maildir_sync_context, 1);
ctx->ibox = ibox;
ctx->new_dir = t_strconcat(ibox->path, "/new", NULL);
ctx->cur_dir = t_strconcat(ibox->path, "/cur", NULL);
return ctx;
}
static void maildir_sync_deinit(struct maildir_sync_context *ctx)
{
if (ctx->uidlist_sync_ctx != NULL)
(void)maildir_uidlist_sync_deinit(ctx->uidlist_sync_ctx);
}
static int maildir_fix_duplicate(struct index_mailbox *ibox, const char *dir,
const char *old_fname)
{
const char *new_fname, *old_path, *new_path;
int ret = 0;
t_push();
old_path = t_strconcat(dir, "/", old_fname, NULL);
new_fname = maildir_generate_tmp_filename(&ioloop_timeval);
new_path = t_strconcat(ibox->path, "/new/", new_fname, NULL);
if (rename(old_path, new_path) == 0) {
i_warning("Fixed duplicate in %s: %s -> %s",
ibox->path, old_fname, new_fname);
} else if (errno != ENOENT) {
mail_storage_set_critical(ibox->box.storage,
"rename(%s, %s) failed: %m", old_path, new_path);
ret = -1;
}
t_pop();
return ret;
}
static int maildir_scan_dir(struct maildir_sync_context *ctx, int new_dir)
{
struct mail_storage *storage = ctx->ibox->box.storage;
const char *dir;
DIR *dirp;
string_t *src, *dest;
struct dirent *dp;
enum maildir_uidlist_rec_flag flags;
int move_new, ret = 1;
src = t_str_new(1024);
dest = t_str_new(1024);
dir = new_dir ? ctx->new_dir : ctx->cur_dir;
dirp = opendir(dir);
if (dirp == NULL) {
mail_storage_set_critical(storage,
"opendir(%s) failed: %m", dir);
return -1;
}
move_new = new_dir && !mailbox_is_readonly(&ctx->ibox->box);
while ((dp = readdir(dirp)) != NULL) {
if (dp->d_name[0] == '.')
continue;
flags = 0;
if (move_new) {
str_truncate(src, 0);
str_truncate(dest, 0);
str_printfa(src, "%s/%s", ctx->new_dir, dp->d_name);
str_printfa(dest, "%s/%s", ctx->cur_dir, dp->d_name);
if (strchr(dp->d_name, ':') == NULL)
str_append(dest, ":2,");
if (rename(str_c(src), str_c(dest)) == 0) {
/* we moved it - it's \Recent for use */
ctx->ibox->dirty_cur_time = ioloop_time;
flags |= MAILDIR_UIDLIST_REC_FLAG_MOVED |
MAILDIR_UIDLIST_REC_FLAG_RECENT;
} else if (ENOTFOUND(errno)) {
/* someone else moved it already */
flags |= MAILDIR_UIDLIST_REC_FLAG_MOVED;
} else if (ENOSPACE(errno)) {
/* not enough disk space, leave here */
flags |= MAILDIR_UIDLIST_REC_FLAG_NEW_DIR |
MAILDIR_UIDLIST_REC_FLAG_RECENT;
move_new = FALSE;
} else {
flags |= MAILDIR_UIDLIST_REC_FLAG_NEW_DIR |
MAILDIR_UIDLIST_REC_FLAG_RECENT;
mail_storage_set_critical(storage,
"rename(%s, %s) failed: %m",
str_c(src), str_c(dest));
}
} else if (new_dir) {
flags |= MAILDIR_UIDLIST_REC_FLAG_NEW_DIR |
MAILDIR_UIDLIST_REC_FLAG_RECENT;
}
ret = maildir_uidlist_sync_next(ctx->uidlist_sync_ctx,
dp->d_name, flags);
if (ret <= 0) {
if (ret < 0)
break;
/* possibly duplicate - try fixing it */
if (maildir_fix_duplicate(ctx->ibox,
dir, dp->d_name) < 0) {
ret = -1;
break;
}
}
}
if (closedir(dirp) < 0) {
mail_storage_set_critical(storage,
"closedir(%s) failed: %m", dir);
}
return ret < 0 ? -1 : 0;
}
static int maildir_sync_quick_check(struct maildir_sync_context *ctx,
int *new_changed_r, int *cur_changed_r)
{
const struct mail_index_header *hdr;
struct index_mailbox *ibox = ctx->ibox;
struct stat st;
time_t new_mtime, cur_mtime;
*new_changed_r = *cur_changed_r = FALSE;
if (stat(ctx->new_dir, &st) < 0) {
mail_storage_set_critical(ibox->box.storage,
"stat(%s) failed: %m", ctx->new_dir);
return -1;
}
new_mtime = st.st_mtime;
if (stat(ctx->cur_dir, &st) < 0) {
mail_storage_set_critical(ibox->box.storage,
"stat(%s) failed: %m", ctx->cur_dir);
return -1;
}
cur_mtime = st.st_mtime;
if (ibox->last_cur_mtime == 0) {
/* first sync in this session, get cur stamp from index */
if (mail_index_get_header(ibox->view, &hdr) == 0)
ibox->last_cur_mtime = hdr->sync_stamp;
}
if (new_mtime != ibox->last_new_mtime ||
new_mtime >= ibox->last_new_sync_time - MAILDIR_SYNC_SECS) {
*new_changed_r = TRUE;
ibox->last_new_mtime = new_mtime;
ibox->last_new_sync_time = ioloop_time;
}
if (cur_mtime != ibox->last_cur_mtime ||
(ibox->dirty_cur_time != 0 &&
ioloop_time - ibox->dirty_cur_time > MAILDIR_SYNC_SECS)) {
/* cur/ changed, or delayed cur/ check */
*cur_changed_r = TRUE;
ibox->last_cur_mtime = cur_mtime;
}
ibox->dirty_cur_time =
cur_mtime >= ioloop_time - MAILDIR_SYNC_SECS ? cur_mtime : 0;
return 0;
}
static int maildir_sync_index(struct maildir_sync_context *ctx)
{
struct index_mailbox *ibox = ctx->ibox;
struct maildir_index_sync_context sync_ctx;
struct maildir_uidlist_iter_ctx *iter;
struct mail_index_transaction *trans;
struct mail_index_view *view;
const struct mail_index_header *hdr;
const struct mail_index_record *rec;
uint32_t seq, uid;
enum maildir_uidlist_rec_flag uflags;
const char *filename;
enum mail_flags flags;
keywords_mask_t keywords;
uint32_t sync_stamp;
int ret;
memset(&sync_ctx, 0, sizeof(sync_ctx));
sync_ctx.ibox = ibox;
if (mail_index_sync_begin(ibox->index, &sync_ctx.sync_ctx, &view,
(uint32_t)-1, (uoff_t)-1) <= 0) {
mail_storage_set_index_error(ibox);
return -1;
}
sync_ctx.view = view;
ret = mail_index_get_header(view, &hdr);
i_assert(ret == 0); /* view is locked, can't happen */
trans = mail_index_transaction_begin(view, FALSE);
seq = 0;
iter = maildir_uidlist_iter_init(ibox->uidlist);
while (maildir_uidlist_iter_next(iter, &uid, &uflags, &filename)) {
maildir_filename_get_flags(filename, &flags, keywords);
__again:
seq++;
if ((uflags & MAILDIR_UIDLIST_REC_FLAG_NONSYNCED) != 0) {
/* partial syncing */
continue;
}
if (seq > hdr->messages_count) {
mail_index_append(trans, uid, &seq);
mail_index_update_flags(trans, seq, MODIFY_REPLACE,
flags, keywords);
continue;
}
if (mail_index_lookup(view, seq, &rec) < 0) {
mail_storage_set_index_error(ibox);
ret = -1;
break;
}
if (rec->uid < uid) {
/* expunged */
mail_index_expunge(trans, seq);
goto __again;
}
if (rec->uid > uid) {
/* new UID in the middle of the mailbox -
shouldn't happen */
mail_storage_set_critical(ibox->box.storage,
"Maildir sync: UID inserted in the middle "
"of mailbox (%u > %u)", rec->uid, uid);
mail_index_mark_corrupted(ibox->index);
ret = -1;
break;
}
if ((rec->flags & MAIL_INDEX_MAIL_FLAG_DIRTY) != 0) {
/* we haven't been able to update maildir with this
record's flag changes. don't sync them. */
continue;
}
maildir_filename_get_flags(filename, &flags, keywords);
if ((uint8_t)flags != (rec->flags & MAIL_FLAGS_MASK) ||
memcmp(keywords, rec->keywords,
INDEX_KEYWORDS_BYTE_COUNT) != 0) {
mail_index_update_flags(trans, seq, MODIFY_REPLACE,
flags, keywords);
}
}
maildir_uidlist_iter_deinit(iter);
if (!ctx->partial) {
/* expunge the rest */
for (seq++; seq <= hdr->messages_count; seq++)
mail_index_expunge(trans, seq);
}
if (ret < 0)
mail_index_transaction_rollback(trans);
else {
uint32_t seq;
uoff_t offset;
if (mail_index_transaction_commit(trans, &seq, &offset) < 0)
mail_storage_set_index_error(ibox);
else {
ibox->commit_log_file_seq = seq;
ibox->commit_log_file_offset = offset;
}
}
/* now, sync the index */
while ((ret = mail_index_sync_next(sync_ctx.sync_ctx,
&sync_ctx.sync_rec)) > 0) {
if (maildir_sync_record(ibox, &sync_ctx) < 0) {
ret = -1;
break;
}
}
sync_stamp = ibox->dirty_cur_time != 0 ? 0 : ibox->last_cur_mtime;
if (mail_index_sync_end(sync_ctx.sync_ctx, sync_stamp, 0) < 0)
ret = -1;
if (ret == 0) {
ibox->commit_log_file_seq = 0;
ibox->commit_log_file_offset = 0;
} else {
mail_storage_set_index_error(ibox);
}
return ret;
}
static int maildir_sync_context(struct maildir_sync_context *ctx)
{
int ret, new_changed, cur_changed;
if (maildir_sync_quick_check(ctx, &new_changed, &cur_changed) < 0)
return -1;
if (!new_changed && !cur_changed)
return 0;
ctx->partial = !cur_changed;
ctx->uidlist_sync_ctx =
maildir_uidlist_sync_init(ctx->ibox->uidlist, ctx->partial);
if (maildir_scan_dir(ctx, TRUE) < 0)
return -1;
if (cur_changed) {
if (maildir_scan_dir(ctx, FALSE) < 0)
return -1;
}
ret = maildir_uidlist_sync_deinit(ctx->uidlist_sync_ctx);
ctx->uidlist_sync_ctx = NULL;
if (ret == 0)
ret = maildir_sync_index(ctx);
return ret;
}
static int maildir_sync_context_readonly(struct maildir_sync_context *ctx)
{
int ret;
ctx->uidlist_sync_ctx =
maildir_uidlist_sync_init(ctx->ibox->uidlist, FALSE);
if (maildir_scan_dir(ctx, TRUE) < 0)
return -1;
if (maildir_scan_dir(ctx, FALSE) < 0)
return -1;
ret = maildir_uidlist_sync_deinit(ctx->uidlist_sync_ctx);
ctx->uidlist_sync_ctx = NULL;
return ret;
}
int maildir_storage_sync_readonly(struct index_mailbox *ibox)
{
struct maildir_sync_context *ctx;
int ret;
ctx = maildir_sync_context_new(ibox);
ret = maildir_sync_context_readonly(ctx);
maildir_sync_deinit(ctx);
return ret;
}
int maildir_storage_sync(struct mailbox *box, enum mailbox_sync_flags flags)
{
struct index_mailbox *ibox = (struct index_mailbox *)box;
struct maildir_sync_context *ctx;
int ret;
if ((flags & MAILBOX_SYNC_FLAG_FAST) == 0 ||
ibox->sync_last_check + MAILBOX_FULL_SYNC_INTERVAL <= ioloop_time) {
ibox->sync_last_check = ioloop_time;
ctx = maildir_sync_context_new(ibox);
ret = maildir_sync_context(ctx);
maildir_sync_deinit(ctx);
if (ret < 0)
return -1;
}
return index_storage_sync(box, flags);
}