journal: if we encounter a corrupted file, rotate and go on

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index 8c17620ed339bf9576bf498e2d6ad02018983df8..1e6d78cfa43606414b4b4d026491d9b1e0bdaf1a 100644 (file)
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -1727,6 +1727,9 @@ int journal_file_open(
             (flags & O_ACCMODE) != O_RDWR)
                 return -EINVAL;
 
+        if (!endswith(fname, ".journal"))
+                return -EINVAL;
+
         f = new0(JournalFile, 1);
         if (!f)
                 return -ENOMEM;
@@ -1840,7 +1843,7 @@ int journal_file_rotate(JournalFile **f) {
 
         l = strlen(old_file->path);
 
-        p = new(char, l + 1 + 16 + 1 + 32 + 1 + 16 + 1);
+        p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1);
         if (!p)
                 return -ENOMEM;
 
@@ -1867,6 +1870,44 @@ int journal_file_rotate(JournalFile **f) {
         return r;
 }
 
+int journal_file_open_reliably(
+                const char *fname,
+                int flags,
+                mode_t mode,
+                JournalFile *template,
+                JournalFile **ret) {
+
+        int r;
+        size_t l;
+        char *p;
+
+        r = journal_file_open(fname, flags, mode, template, ret);
+        if (r != -EBADMSG)
+                return r;
+
+        if ((flags & O_ACCMODE) == O_RDONLY)
+                return r;
+
+        if (!(flags & O_CREAT))
+                return r;
+
+        l = strlen(fname);
+        if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
+                     (int) (l-8), fname,
+                     (unsigned long long) now(CLOCK_REALTIME),
+                     random_ull()) < 0)
+                return -ENOMEM;
+
+        r = rename(fname, p);
+        free(p);
+        if (r < 0)
+                return -errno;
+
+        log_warning("File %s corrupted, renaming and replacing.", fname);
+
+        return journal_file_open(fname, flags, mode, template, ret);
+}
+
 struct vacuum_info {
         off_t usage;
         char *filename;

diff --git a/src/journal/journal-file.h b/src/journal/journal-file.h
index 4ef4a14479e88ee023f1c46f5c0fdd766d55c413..4d6e1407711bad50df1d7896f88c0399cbd8df24 100644 (file)
--- a/src/journal/journal-file.h
+++ b/src/journal/journal-file.h
@@ -89,6 +89,8 @@ typedef enum direction {
 int journal_file_open(const char *fname, int flags, mode_t mode, JournalFile *template, JournalFile **ret);
 void journal_file_close(JournalFile *j);
 
+int journal_file_open_reliably(const char *fname, int flags, mode_t mode, JournalFile *template, JournalFile **ret);
+
 int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret);
 
 uint64_t journal_file_entry_n_items(Object *o);

diff --git a/src/journal/journald.c b/src/journal/journald.c
index 37e6484f4b1d2442b5f3bae69eef81e284da7730..e9ac897de7fd37793f5fa654274ebf5ed65e063b 100644 (file)
--- a/src/journal/journald.c
+++ b/src/journal/journald.c
@@ -301,7 +301,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) {
                 journal_file_close(f);
         }
 
-        r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
+        r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
         free(p);
 
         if (r < 0)
@@ -604,8 +604,12 @@ retry:
         else {
                 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
 
-                if (r == -E2BIG && !vacuumed) {
-                        log_info("Allocation limit reached.");
+                if ((r == -EBADMSG || r == -E2BIG) && !vacuumed) {
+
+                        if (r == -E2BIG)
+                                log_info("Allocation limit reached, rotating.");
+                        else
+                                log_warning("Journal file corrupted, rotating.");
 
                         server_rotate(s);
                         server_vacuum(s);
@@ -1875,7 +1879,7 @@ static int system_journal_open(Server *s) {
                 if (!fn)
                         return -ENOMEM;
 
-                r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
+                r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
                 free(fn);
 
                 if (r >= 0) {
@@ -1906,7 +1910,7 @@ static int system_journal_open(Server *s) {
                          * if it already exists, so that we can flush
                          * it into the system journal */
 
-                        r = journal_file_open(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
+                        r = journal_file_open_reliably(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
                         free(fn);
 
                         if (r < 0) {
@@ -1922,7 +1926,7 @@ static int system_journal_open(Server *s) {
                          * it if necessary. */
 
                         (void) mkdir_parents(fn, 0755);
-                        r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
+                        r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
                         free(fn);
 
                         if (r < 0) {
@@ -2666,10 +2670,6 @@ static int server_init(Server *s) {
         if (r < 0)
                 return r;
 
-        r = system_journal_open(s);
-        if (r < 0)
-                return r;
-
         r = open_signalfd(s);
         if (r < 0)
                 return r;
@@ -2678,6 +2678,10 @@ static int server_init(Server *s) {
         if (!s->rate_limit)
                 return -ENOMEM;
 
+        r = system_journal_open(s);
+        if (r < 0)
+                return r;
+
         return 0;
 }