Fix potential EIO error #51
Conversation
mkfs.c
Outdated
| stat_buf.st_size = blk_size; | ||
| } | ||
|
|
||
| /* Initialize the file system image with zeros */ |
There was a problem hiding this comment.
You should clear a root index block instead. That is, calloc(SIMPLEFS_BLOCK_SIZE) and write(fd, block, SIMPLEFS_BLOCK_SIZE). In addition, do verify the result of write system call.
There was a problem hiding this comment.
Hi @jserv, I think this issue is related to #20. The problem arises from incorrect reading of simplefs_file_ei_block. If the values inside blockdata are random or not the correct format, then the resulting simplefs_file_ei_block will be incorrect, causing the initial nr_files and simplefs_extent to have meaningless values. However, this leads to an sb_bread error during simplefs_lookup and other directory related problems.
To address this issue, the solution is to initialize the file system image with zeros instead of just clearing a root index block. This ensures that the values within blockdata are consistent and prevent the occurrence of random values.
Alternatively, should I approach this issue differently to find a solution?
There was a problem hiding this comment.
To address this issue, the solution is to initialize the file system image with zeros instead of just clearing a root index block. This ensures that the values within blockdata are consistent and prevent the occurrence of random values. Alternatively, should I approach this issue differently to find a solution?
I would expect the minimal operations against the raw file system image. That is, to clear a root index block instead of zeroing the whole file system image. You should investigate the necessary changes required to make the file system consistent.
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
If we use `dd if=/dev/random of=test.img bs=1M count=32` rather then `if=/dev/zero` and mount it, it will cause EIO error and stuck in for loop. This commit addresses an issue where `ls` commands would read random , meaningless values from `simplefs_file_ei_block` for uninitialized blocks. This occurred because the blocks contained residual data, leading to erroneous outputs. Changes: 1. Clear the root extent block during the `mkfs` process to prevent error `simplefs_file_ei_block` data. 2. Ensured that file data blocks are cleared immediately upon allocation to prevent error `simplefs_file` data. Impact: By clearing these blocks, we prevent the mistake during `ls` command and resolve Issue sysprog21#51 and sysprog21#20.
bitmap.h
Outdated
| struct simplefs_sb_info *sbi = SIMPLEFS_SB(sb); | ||
| uint32_t ret = get_first_free_bits(sbi->bfree_bitmap, sbi->nr_blocks, len); | ||
| if (ret) | ||
| if (ret) { |
There was a problem hiding this comment.
Replace with the following:
if (!ret) /* Add proper comments here */
return 0;
sbi->nr_free_blocks -= len;
...
mkfs.c
Outdated
| static int write_data_blocks(int fd, struct superblock *sb) | ||
| { | ||
| /* FIXME: unimplemented */ | ||
| char *buffer = (char *) calloc(1, SIMPLEFS_BLOCK_SIZE); |
jserv
left a comment
jserv
left a comment
There was a problem hiding this comment.
Append Close #20 at the end of git commit message.
If we use 'dd if=/dev/random of=test.img bs=1M count=32' rather then 'if=/dev/zero' and mount it, it will cause EIO error and stuck in for loop. This commit addresses an issue where 'ls' commands would read random , meaningless values from 'simplefs_file_ei_block' for uninitialized blocks. This occurred because the blocks contained residual data, leading to erroneous outputs. Changes: 1. Clear the root extent block during the 'mkfs' process to prevent error 'simplefs_file_ei_block' data. 2. Ensured that file data blocks are cleared immediately upon allocation to prevent error 'simplefs_file' data. Impact: By clearing these blocks, we prevent the mistake during 'ls' command and resolve Issue sysprog21#51 and sysprog21#20. Close sysprog21#20 and sysprog21#51
|
Thank @HotMercury for contributing! |
2 participants
This PR addressed issue #50 , but I don't think I fully understand the reason why that if we not initial the disk to zero, it will cause error happended.