2010-09-06 21:58:36 process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93198
hi :
when i writed files into the harddisk , my process was killed. it seen that there was no enough memory to be used. my test code as following :
#define PTH_NUM 32
static unsigned int pth_para[PTH_NUM];
static unsigned int file_num[PTH_NUM];
void * pthread_fun (void * para)
{
FILE * fd;
int count;
int len;
char buf[128];
unsigned int id;
if (! para) return NULL;
id = * (unsigned int *) para;
printf("pthread id : %02d\n", id);
if(id >= PTH_NUM) return NULL;
file_num[id] = 0;
while(1) {
snprintf(buf, 128, "/mnt/%02d_test%06d", id, file_num[id]);
file_num[id]++;
fd = fopen(buf, "wb");
if (! fd) {
perror("open file");
return NULL;
}
count = 0;
snprintf(buf, 128, "test data | ");
len = strlen(buf);
rewind(fd);
while (1) {
fwrite(buf, 1, len, fd);
count += len;
if((count % 120) == 0) {
fprintf(fd, "\n");
}
if (count >= 64 * 1024 * 1024) {
break;
}
}
fclose(fd);
}
return 0;
}
int main (int argc, char **argv)
{
int i;
pthread_t tid[PTH_NUM];
for (i = 0; i < PTH_NUM; i++) {
pth_para[i] = i;
pthread_create (&tid[i], NULL, pthread_fun, &pth_para[i]);
}
while(1) {
sleep(5);
}
return 0;
}
i don't use the dynamic memory , why the memory was lost too much? is it writing data too fast here????
QuoteReplyEditDelete
2010-09-07 00:18:36 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93199
Which blackfin board do you use? What is the error log when the process is killed?
QuoteReplyEditDelete
2010-09-07 01:05:44 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93201
BF533-EZKIT, 2008R1, no error information was printed. the hard disk indicator is always in the light 。so i think the hardisk not work well except reseting. and all the command in shell can't run any more, they were killed at once.
QuoteReplyEditDelete
2010-09-07 01:52:15 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93202
Could you update to 2009R1.1? 2008R1 is out of date.
If you run all thread in sequence, you may avoid the out of memory issue. Otherwise, 64M bytes * 32 files * 32 threads consume too much VFS buffers concurrently.
QuoteReplyEditDelete
2010-09-08 00:22:23 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93229
i test it last night again, and printing following message now.
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
end_request: I/O error, dev hda, sector 151
end_request: I/O error, dev hda, sector 63
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
end_request: I/O error, dev hda, sector 63
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 0, block_bitmap = 11
end_request: I/O error, dev hda, sector 63
it seens that the hardware had something wrong, is it right?? can you give me some advice?
QuoteReplyEditDelete
2010-09-08 02:10:38 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93234
This doesn't look like a out of memory issue. Check your disk driver.
QuoteReplyEditDelete
2010-09-08 07:56:37 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93260
where can download the latest driver ?
by the way , why my browser can't display the post button?????
QuoteReplyEditDelete
2010-09-09 00:09:41 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93276
Use firefox other than IE
QuoteReplyEditDelete
2010-09-09 10:13:49 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93302
the ide driver doesn't support the samsung hardisk, the other hardisk works well. i don't know the difference between them.
can you give me some information ?
QuoteReplyEditDelete
2010-09-13 09:38:10 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93383
now, if i do nothing, the process writing the hadisk runs well. but if there are many documents in de harddisk and i do the "ls" command, the writing process was killed sometimes. i know the "ls" command needs a lots of memory and the memory is very nervous, so the system will invoke the swap process . i just have a question, this case is normal? if not, what should i do? thanks.
QuoteReplyEditDelete
2010-09-14 00:32:07 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93396
Don't write so many files concurrently or don't run ls at the same time.
QuoteReplyEditDelete
2010-09-15 23:26:00 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93479
does 2008r1-8 toolchain support the large file option in default ??
QuoteReplyEditDelete
2010-09-16 00:17:48 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93481
large file support is included in uClibc full rpm package. Which uClibc rpm did you install?
In addition, 2008R1 is not supported any more. Please upgrade to 2009R1.1.
QuoteReplyEditDelete
2010-09-16 00:28:04 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93482
i am using blackfin-toolchain-uclibc-full-08rl-8 now , i just want to upgrade the toolchain to 2009R1.1, the kernel also use the 2008R1, is it OK?
QuoteReplyEditDelete
2010-09-16 01:31:05 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93483
No. See docs.blackfin.uclinux.org/doku.php?id=faq
QuoteReplyEditDelete
2010-09-16 21:44:45 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93517
i can't open the link. i need your help, please give me some advice.
QuoteReplyEditDelete
2010-09-17 00:20:11 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93521
The link says you should use the same toolchain and Linux distribution release. Such as 2009R1.1.
QuoteReplyEditDelete
2010-09-17 22:17:03 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93568
Hi , sonic
i catched the following message by using "dmesg" command :
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 419, block_bitmap = 13729803
end_request: I/O error, dev hda, sector 109838487
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 419, block_bitmap = 13729803
end_request: I/O error, dev hda, sector 109838487
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 419, block_bitmap = 13729803
end_request: I/O error, dev hda, sector 109838487
end_request: I/O error, dev hda, sector 63
EXT2-fs error (device hda1): read_block_bitmap: Cannot read block bitmap - block_group = 419, block_bitmap = 13729803
Out of memory: kill process 213 (sport_test) score 45 or a child
Killed process 213 (sport_test)
Out of memory: kill process 208 (sh) score 13 or a child
Killed process 208 (sh)
end_request: I/O error, dev hda, sector 109576367
end_request: I/O error, dev hda, sector 63
which case will lead to invoke breath function failed ? the problem puzzle me long time, please help me, thanks
QuoteReplyEditDelete
2010-09-17 22:20:15 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93569
Sorry, thar is bread function.
QuoteReplyEditDelete
2010-09-17 22:24:57 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93570
read_block_bitmap sb_bread function
QuoteReplyEditDelete
2010-09-20 02:18:18 Re: process was killed when writing the hardisk
Sonic Zhang (CHINA)
Message: 93624
May be problem in your disk driver. Try the same operation on a USB storage device first.
QuoteReplyEditDelete
2010-09-21 03:34:06 Re: process was killed when writing the hardisk
steven O (AFGHANISTAN)
Message: 93693
i use the mke2fs to format the /dev/hda1, if the space of hda1 less than 2G. the operation run successfully. if not, the operation failed. it seems that the uClibc don't support the LFS option. but the uClibc spportted LFS option in default. why ?