-
Notifications
You must be signed in to change notification settings - Fork 73
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improve error position detention at WRITE PERM #357
Improve error position detention at WRITE PERM #357
Conversation
Ready for review |
Cherry picked to the |
* Correct miscalculation of last block on tape * Consider the final index on the partition as error position even if very small number is returned * Never adjust the force_writeperm threshold for better debug * Stop checking the I/F of the ltfs-backends repository
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Code looks good to me with additional validations, a couple I do not fully understand but questions/comments documented for Abe-san who does know the code (I am just learning through it)
if (last_index_pos > err_pos.block) { | ||
ltfsmsg(LTFS_INFO, 13027I, (int)err_pos.partition, | ||
(unsigned long long)err_pos.block, (unsigned long long)last_index_pos); | ||
err_pos.block = last_index_pos + 1; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@piste-jp-ibm So if last_index_pos is "the end of the partition" you are moving error_pos to the end of the partition?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is first block of the index. But it is enough because every extents must not have that block at all.
So LTFS sets the last position on tape to last_index_pos + 1
and cleanup extents. As a result, all extents after the last index will be removed.
memcpy(pos, &dev->position, sizeof(struct tc_position)); | ||
|
||
ltfsmsg(LTFS_DEBUG, 11335D, (int)pos->block, block); | ||
pos->block -= block; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@piste-jp-ibm at first glance I would have thought on this line as the guilty of the overwriting of the previous block
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes. this line is one of the root cause. But logic itself is correct. The idea how to fetch the error position is bad.
We thought that the only way to fetch the error position is subtract number of records in buffer from the current position. But this idea is not good when a WRITE PERM happens just after a locate (for append). Because the drive mould have records which is read by the locate in buffer.
Actually, we realized error position it self is provided by READ_POSITION command.
if (ext->start.block && ext->bytecount) { | ||
extent_last.partition = ltfs_part_id2num(ext->start.partition, vol); | ||
/* Calculate the last block of this extent */ | ||
extent_last.block = ext->start.block + (ext->bytecount / blocksize); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@piste-jp-ibm so LTFS always uses fixed length never variable length right?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No, a extent is constructed from one or more 512KB fixed blocks and no or one variable block.
Summary of changes
This pull request includes following changes or fixes.
Description
Fixes #356
Type of change
Checklist:
Test Log