Skip to contentSkip to conversation

fs/growfiles.c: Fix the max size of every growing file

PATCH New Xiao Yang 2 participants 4 messages Oct 17, 2018 – Dec 4, 2018

lore
Share & tools
lore

Xiao Yang

author
#0 lore raw

PATCH 1/1 fs/growfiles.c: Fix the max size of every growing file

If -i option is set to zero, some growfiles tests (e.g. gf01, gf[10-11], gf[14-15], gf18) will grow every file size to 2G before stopping test. On less 2G free space, these growfiles tests fails with ENOSPC early because they cannot grow size to 2G for every file.

We try to compare actual max size of every growing file with default size(i.e. 2G) and use the samller one as the limit, so that growing size for every file doesn't fail with ENOSPC on less 2G free space.

Note:
-T/-U option will truncate or remove some files within specified interval, but we still give enough space for every file to grow. we still restrict the max size of every growing file to less than 2G.

Fix: #389

---
 testcases/kernel/fs/doio/growfiles.c | 21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)
testcases/kernel/fs/doio/growfiles.c +17 -4
index 02df1b28c..eb58ce0cd 100644
@@ -78,6 +78,8 @@
 #include <sys/time.h>
 #include <sys/param.h>
 #include <sys/signal.h>
+#include <sys/statfs.h>
+#include <sys/vfs.h>
 #include <fcntl.h>
 #include <errno.h>
 #include <string.h>
@@ -367,7 +369,8 @@ int main(int argc, char **argv)
 	int stop = 0;		/* loop stopper if set */
 
 	unsigned long curr_size = 0;	/* BUG:14136 (keep track of file size) */
-	const unsigned long ext2_limit = 2147483647;	/* BUG:14136 (2GB ext2 filesize limit) */
+	unsigned long fs_limit = 2147483647; /* BUG:14136 (filesystem size limit is 2G by default) */
+	struct statfs fsbuf;
 
 	int tmp;
 	char chr;
@@ -1360,6 +1363,16 @@ no whole file checking will be performed!\n", Progname, TagName,
 #endif
 	}
 
+	if (statfs(auto_dir, &fsbuf) == -1) {
+		fprintf(stderr, "%s%s: Unable to get the info of mounted "
+			"filesystem that includes dir %s\n",
+			Progname, TagName, auto_dir);
+		exit(1);
+	}
+
+	/* Compare two values and use the smaller one as limit */
+	fs_limit = MIN(fsbuf.f_bsize * fsbuf.f_bavail / num_files, fs_limit);
+
 	/*
 	 * This is the main iteration loop.
 	 * Each iteration, all files can  be opened, written to,
@@ -1476,14 +1489,14 @@ no whole file checking will be performed!\n", Progname, TagName,
 			 * if we are dealing with a FIFO file.
 			 */
 
-			/* BUG:14136 (don't go past ext2's filesize limit) */
+			/* BUG:14136 (don't go past filesystem size limit) */
 			curr_size = file_size(fd);
-			if (curr_size + grow_incr >= ext2_limit) {
+			if (curr_size + grow_incr >= fs_limit) {
 				lkfile(fd, LOCK_UN, LKLVL1);	/* release lock */
 				close(fd);
 				sprintf(reason,
 					"Reached %ld filesize which is almost %ld limit.",
-					curr_size, ext2_limit);
+					curr_size, fs_limit);
 				stop = 1;
 				continue;
 			}
Signature · 1 line
2.17.0
#1 lore raw

PATCH 1/1 fs/growfiles.c: Fix the max size of every growing file

Hi Xiao,

...

testcases/kernel/fs/doio/growfiles.c int main(int argc, char **argv) · line 372

Quoted 3 lines
testcases/kernel/fs/doio/growfiles.c +1 -1
 	unsigned long curr_size = 0;	/* BUG:14136 (keep track of file size) */
-	const unsigned long ext2_limit = 2147483647;	/* BUG:14136 (2GB ext2 filesize limit) */
+	unsigned long fs_limit = 2147483647; /* BUG:14136 (filesystem size limit is 2G by default) */

BTW I wonder if the "bug" is this one:
https://access.redhat.com/solutions/29129

doio code really needs cleanup.

Kind regards,
Petr

author
#2 lore raw

PATCH 1/1 fs/growfiles.c: Fix the max size of every growing file

Quoted 10 lines · Petr Vorel

Hi Xiao,

...

testcases/kernel/fs/doio/growfiles.c int main(int argc, char **argv) · line 372

Quoted 3 lines

unsigned long curr_size = 0; /* BUG:14136 (keep track of file size) */

testcases/kernel/fs/doio/growfiles.c +1 -1
-	const unsigned long ext2_limit = 2147483647;	/* BUG:14136 (2GB ext2 filesize limit) */
+	unsigned long fs_limit = 2147483647; /* BUG:14136 (filesystem size limit is 2G by default) */

BTW I wonder if the "bug" is this one:
https://access.redhat.com/solutions/29129

Hi Petr,

Sorry, i am not sure.

The following old patch try to fix defect 14136, but i don't know what the defect 14136 is.
https://github.com/linux-test-project/ltp/commit/dec5a46e9939ac99c30afb5b2605dbd9aa9ed6ef

I cannot find the bug(id: 14136) of filesystem, is it the issue of LTP?

Quoted 1 line

doio code really needs cleanup.

Agreed.

Best Regards,
Xiao Yang

Quoted 5 lines

Kind regards,
Petr