refactor: catch errno EINTR when reading and writing files instead of EAGAIN

[guoxiangCN]

EAGAIN error code Appears only when the O_NONBLOCK option is set and there is no more content in the file to read.
If a read() is interrupted by a signal before it reads any data, it shall return -1 with errno set to [EINTR].

[codecov]

Codecov Report

Merging #242 (d2b41e0) into master (ee0f6cf) will increase coverage by 0.01%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #242      +/-   ##
==========================================
+ Coverage   97.18%   97.19%   +0.01%     
==========================================
  Files          30       30              
  Lines        9368     9368              
==========================================
+ Hits         9104     9105       +1     
+ Misses        264      263       -1     

Impacted Files	Coverage Δ
src/env/default.rs	92.13% <0.00%> (ø)
src/codec.rs	98.67% <0.00%> (+0.33%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ee0f6cf...d2b41e0. Read the comment docs.

[tabokie]

Thanks for the attention.

According to rust, a Write::write shouldn't write any bytes if an error is returned:

If an error is returned then no bytes in the buffer were written to this writer.

So it appears we should remove the retry looping entirely. For consistency do the same for the read as well.

FYI we already have retry logic at a higher level, so this change won't affect the IO results.

raft-engine/src/file_pipe_log/log_file.rs

Line 203 in ee0f6cf

Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => {}

raft-engine/src/file_pipe_log/log_file.rs

Line 107 in ee0f6cf

self.writer.write_all(buf)?;

[LykxSassinator]

How about direclty add an extra option OFlag::O_NONBLOCK for the open api rather than modify the Errno ? I think the original design on it is to use OFlag::O_NONBLOCK as the default option for fd.

[guoxiangCN]

Okay, I didn't notice the higher judgment. However, according to the constraints of the write method you mentioned, retrying the entire while loop will also lead to a semi successful and semi failed state of data writing, that is, partial success and partial failure.
However, I'm not sure whether I need to cross the loop to really write. I don't know much about the underlying file system writing.

[tabokie]

Having looked closely, I think the implementation here is not very important. We maintain a file offset and use it to write, so a partial successful write can be overwritten correctly afterwards:

raft-engine/src/env/default.rs

Lines 215 to 216 in d2b41e0

	let len = self.inner.write(self.offset, buf)?;
	self.offset += len;

And the pwrite interface doesn't guarantee atomicity at all. (In fact non of the unix syscalls guarantees that, considering direct IO). So your current changes look okay.

[tabokie]

I think the original design on it is to use OFlag::O_NONBLOCK as the default option for fd.

@LykxSassinator No, non block is only useful for network fd.