Adds the new -L, --error-limit options.

And also sets the default limit to 100.

HISTORY

@@ -3,6 +3,13 @@ SnapRAID HISTORY
 
 8.0 2015/01 BETA
 ================
+ * Allows "sync" and "scrub" to continue after the first
+   bunch of disk errors. Blocks with errors are marked as bad,
+   and you can try to fix them with the "fix -e" command.
+   The fix should then trigger the internal disk firmware to
+   reallocate the bad sector, and likely fixing the problem.
+   You can control the number of allowed errors with the new
+   -L, --error-limit option. The default is 100.
  * The -e, --filter-error option now doesn't write anymore fixes
    to unsynced files. This helps in case you are running it on a
    not synced array, removing the risk to revert some files to an

TODO

@@ -5,6 +5,39 @@ This is the list of TODO items for SnapRAID.
 
 - Next
 
+* Improves the fix -e behaviour. At now it reads too mutch data,
+as it needs to verify the full files.
+This is done just to have confirmation that the file is OK,
+and then restore the file timestamp. But other approaches are possible.
+Indicatively we can do:
+- Read/Write only the marked bad error blocks
+- For all the files, owning this block:
+  - If the file is present and it has the original timestamp
+    - If yes, Write any fix to it, and keep the timestamp.
+    - If no, don't write the fix, and assume it as update.
+The issue could be that if the error is in a modified file
+after sync, it won't be fixed. But at now we detect only silent error,
+and they CANNOT be in modified files.
+So, if a file is modified, we just don't write it, at the next scrub
+the bad flag will be removed.
+
+Note that this way of checking the timestamp before writing fixes
+may have sense also for a generic fix.
+We already have the -m, --missing flag. It's already working in this way.
+--missing fixes only deleted/missing files.
+Like --stable fixes only synced files, but unsure of the use case.
+
+We have now:
+- Lost disk. fix -d DISK
+- Deleted files. fix -m / fix -f
+- Silent errors on scrub. fix -e
+Possible new:
+- Disk with failed read, that can be rewritten.
+  - Scrub could put them in error list, and fix could rewrite the sectors to make them right
+    This won't work for modified files, in that case the error will remain.
+    But scrub can report it in the screen and log about the file.
+  - Sync could proceed anyway, just skip such blocks index, and mark the sector as bad.
+
 * Avoid to do optimized move operation if more files with
 same timestamp/size are present.
 

cmdline/snapraid.c

@@ -247,6 +247,7 @@ struct option long_options[] = {
 	{ "older-than", 1, 0, 'o' },
 	{ "start", 1, 0, 'S' },
 	{ "count", 1, 0, 'B' },
+	{ "error-limit", 1, 0, 'L' },
 	{ "import", 1, 0, 'i' },
 	{ "log", 1, 0, 'l' },
 	{ "force-zero", 0, 0, 'Z' },
@@ -257,11 +258,11 @@ struct option long_options[] = {
 	{ "force-full", 0, 0, 'F' },
 	{ "audit-only", 0, 0, 'a' },
 	{ "pre-hash", 0, 0, 'h' },
-	{ "speed-test", 0, 0, 'T' },
+	{ "speed-test", 0, 0, 'T' }, /* undocumented speed test command */
 	{ "gen-conf", 1, 0, 'C' },
 	{ "verbose", 0, 0, 'v' },
-	{ "quiet", 0, 0, 'q' },
-	{ "gui", 0, 0, 'G' }, /* undocumented GUI interface command */
+	{ "quiet", 0, 0, 'q' }, /* undocumented quiet option */
+	{ "gui", 0, 0, 'G' }, /* undocumented GUI interface option */
 	{ "help", 0, 0, 'H' },
 	{ "version", 0, 0, 'V' },
 
@@ -349,7 +350,7 @@ struct option long_options[] = {
 };
 #endif
 
-#define OPTIONS "c:f:d:mep:o:S:B:i:l:ZEUDNFahTC:vqHVG"
+#define OPTIONS "c:f:d:mep:o:S:B:L:i:l:ZEUDNFahTC:vqHVG"
 
 volatile int global_interrupt = 0;
 
@@ -411,6 +412,7 @@ int main(int argc, char* argv[])
 	/* defaults */
 	config(conf, sizeof(conf), argv[0]);
 	memset(&opt, 0, sizeof(opt));
+	opt.io_error_limit = 100;
 	blockstart = 0;
 	blockcount = 0;
 	tommy_list_init(&filterlist_file);
@@ -506,6 +508,15 @@ int main(int argc, char* argv[])
 				/* LCOV_EXCL_STOP */
 			}
 			break;
+		case 'L' :
+			opt.io_error_limit = strtoul(optarg, &e, 0);
+			if (!e || *e) {
+				/* LCOV_EXCL_START */
+				fprintf(stderr, "Invalid error limit number '%s'\n", optarg);
+				exit(EXIT_FAILURE);
+				/* LCOV_EXCL_STOP */
+			}
+			break;
 		case 'i' :
 			if (import_timestamp) {
 				/* LCOV_EXCL_START */

snapraid.1

@@ -21,6 +21,10 @@ snapraid [\-c, \-\-conf CONFIG]
 	[\-p, \-\-percentage PERC] [\-o, \-\-older\-than DAYS]
 .PD 0
 .PP
+.PD
+	[\-l, \-\-log FILE]
+.PD 0
+.PP
 .PD
 	[\-Z, \-\-force\-zero] [\-E, \-\-force\-empty]
 .PD 0
@@ -34,11 +38,15 @@ snapraid [\-c, \-\-conf CONFIG]
 .PD 0
 .PP
 .PD
-	[\-v, \-\-verbose] [\-l, \-\-log FILE]
+	[\-S, \-\-start BLKSTART] [\-B, \-\-count BLKCOUNT]
 .PD 0
 .PP
 .PD
-	[\-S, \-\-start BLKSTART] [\-B, \-\-count BLKCOUNT]
+	[\-L, \-\-error\-limit NUMBER]
+.PD 0
+.PP
+.PD
+	[\-v, \-\-verbose]
 .PD 0
 .PP
 .PD
@@ -871,6 +879,16 @@ to enclose the name in \[dq], like \[dq]>>result.log\[dq].
 To output the log to standard output or standard error,
 you can use respectively \[dq]>&1\[dq] and \[dq]>&2\[dq].
 .TP
+.B \-L, \-\-error\-limit
+Sets a new error limit before stopping execution.
+By default SnapRAID stops if it encouters more than 100
+Input/Output errors, meaning that likely a disk is going to
+die.
+This options affects \[dq]sync\[dq] and \[dq]scrub\[dq], that are allowed
+to continue after the first bunch of disk errors, to try
+to complete at most their operations.
+Instead, \[dq]check\[dq] and \[dq]fix\[dq] always stop at the first error.
+.TP
 .B \-S, \-\-start BLKSTART
 Starts the processing from the specified
 block number. It could be useful to retry to check

snapraid.d

@@ -7,11 +7,13 @@ Synopsis
 	:	[-m, --filter-missing] [-e, --filter-error]
 	:	[-a, --audit-only] [-h, --pre-hash] [-i, --import DIR]
 	:	[-p, --percentage PERC] [-o, --older-than DAYS]
+	:	[-l, --log FILE]
 	:	[-Z, --force-zero] [-E, --force-empty]
 	:	[-U, --force-uuid] [-D, --force-device]
 	:	[-N, --force-nocopy] [-F, --force-full]
-	:	[-v, --verbose] [-l, --log FILE]
 	:	[-S, --start BLKSTART] [-B, --count BLKCOUNT]
+	:	[-L, --error-limit NUMBER]
+	:	[-v, --verbose]
 	:	status|sync|scrub|fix|check|list|diff|dup|pool|rehash
 
 	:snapraid [-V, --version] [-H, --help] [-C, --gen-conf CONTENT]
@@ -634,6 +636,16 @@ Options
 		To output the log to standard output or standard error,
 		you can use respectively ">&1" and ">&2".
 
+	-L, --error-limit
+		Sets a new error limit before stopping execution.
+		By default SnapRAID stops if it encouters more than 100
+		Input/Output errors, meaning that likely a disk is going to
+		die.
+		This options affects "sync" and "scrub", that are allowed
+		to continue after the first bunch of disk errors, to try
+		to complete at most their operations.
+		Instead, "check" and "fix" always stop at the first error.
+
 	-S, --start BLKSTART
 		Starts the processing from the specified
 		block number. It could be useful to retry to check

snapraid.txt

@@ -11,11 +11,13 @@ snapraid [-c, --conf CONFIG]
 	[-m, --filter-missing] [-e, --filter-error]
 	[-a, --audit-only] [-h, --pre-hash] [-i, --import DIR]
 	[-p, --percentage PERC] [-o, --older-than DAYS]
+	[-l, --log FILE]
 	[-Z, --force-zero] [-E, --force-empty]
 	[-U, --force-uuid] [-D, --force-device]
 	[-N, --force-nocopy] [-F, --force-full]
-	[-v, --verbose] [-l, --log FILE]
 	[-S, --start BLKSTART] [-B, --count BLKCOUNT]
+	[-L, --error-limit NUMBER]
+	[-v, --verbose]
 	status|sync|scrub|fix|check|list|diff|dup|pool|rehash
 
 snapraid [-V, --version] [-H, --help] [-C, --gen-conf CONTENT]
@@ -679,6 +681,16 @@ SnapRAID provides the following options:
         To output the log to standard output or standard error,
         you can use respectively ">&1" and ">&2".
 
+    -L, --error-limit
+        Sets a new error limit before stopping execution.
+        By default SnapRAID stops if it encouters more than 100
+        Input/Output errors, meaning that likely a disk is going to
+        die.
+        This options affects "sync" and "scrub", that are allowed
+        to continue after the first bunch of disk errors, to try
+        to complete at most their operations.
+        Instead, "check" and "fix" always stop at the first error.
+
     -S, --start BLKSTART
         Starts the processing from the specified
         block number. It could be useful to retry to check