[VDO-5752] Add vdoformat functionality into the kernel #208
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit adds the parameters that are currently part of vdoformat to the vdo table line. There are three parameters; slab bits, index memory size, and whether the index is sparse or not. The three parameters will be optional and use defaults if not passed in. Additional code has been added to do value checking of these new parameter so we can return errors when values are incorrect. The target_type version number has been updated so that clients of VDO (LVM, Stratis, etc) will know what versions of VDO support these new parameters and the new formatting code. Signed-off-by: Bruce Johnston <[email protected]>
[VDOSTORY-292] This commit provides tests for the new table line parameters that will support formatting in the kernel. It also fixes an issue with the old VDOFormat.pm test which was succeeding even when it shouldn't have been. The first change is to add a flag for deciding what type of formatting we want to do; the old way calling vdoformat and the new way formatting directly from the kernel. Then we create a new function in Unmanaged.pm to do the formatting in the kernel by wiping the first block of the storage and then creating a new VDO volume if parameters were passed into the function. We only do it for testing. If nothing is passed in for testing, the normal dmsetup create done in activate will be used. Then we add a new test; VDOFormatInKernel.pm to test the parameters and validation code. This test is nearly identical to VDOFormat.pm, but has different code for checking error codes and things since they get returned from dmsetup and the specific error messages are written to the kernel log. Finally, some changes are made to VDOFormat.pm. During testing, it was noticed that any error message was accepted because assertExecuteCommand adds the stack trace to the error message and part of the stack trace is the call to tryIllegal with the error message match string. So we changed the search string to be more explicity by adding vdoformat to the prefix. This makes the tests only look at the output of vdoformat. Signed-off-by: Bruce Johnston <[email protected]>
This commit adds a check for an empty geometry block. An empty geometry block tells the code that the vdo device needs to be formatted. This follows similar code in other dm devices like dm-thin. To do this, we split the current read_geometry_block() function into two functions. The first is still called read_geometry_block() but only does the read of the block from disk and does not validate the data. The second is called load_geometry_block and does both the read and validation, like the old read_geometry_block did. The new read_geoemtry_block is now used to read the geometry block and check to see if it is empty. If it is all zeroes, we write a log message out to indicate the vdo needs to be formatted. The actual formatting code will be done in a future commit. Signed-off-by: Bruce Johnston <[email protected]>
[VDOSTORY-292] This commit adds testing support for the previous commit by looking for the log message that is being written when the code determines that formatting is required. Signed-off-by: Bruce Johnston <[email protected]>
This is the actual formatting code. It has been moved from vdoformat into various locations in the kernel code where it makes sense. The heart of the code resides in vdo.c, where we check the result of the previous commit that added a check for empty geometry block. If the block is all zeroes, there is now a call to vdo_format() which will write out the formatting layout for the geometry and super blocks of VDO. Signed-off-by: Bruce Johnston <[email protected]>
This commit adds a user mode version of blkdev_issue_zeroout, which is used in the kernel formatting code to zero out blocks of storage. Signed-off-by: Bruce Johnston <[email protected]>
This commit updates the tests after the formatting code has been added. Changes included disabling certain checks that don't work due to machine restrictions and other things. It also changes how we remove the device after creation by waiting for the index to start before calling dmsetup remove. Signed-off-by: Bruce Johnston <[email protected]>
github-actions
bot
added
public testing running
public testing failed
and removed
untested
public testing running
labels
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This series of commits moves the functionality of vdoformat into the kernel. It includes three main code changes that will go upstream.
Adding three new optional parameters to the vdo table line (slab bits, index memory size, and index sparse) from the vdoformat parameter list, along with validation code and default values for each one.
Adding a check at dmsetup create time to check the geometry block to see if it is all zeroes. This will inform the formatting code whether it should run or not. Basically if the geometry block is all zeroes, we format the device.
The actual formatting code. This is the most detailed of the three main code commits. It will initialize structures with appropriate values so that both the geometry block and geoemtry block can be written out with the correct format. The formatting code will also zero out several areas of the device (uds super block, component areas of the device) so no old info is picked up when it shouldn't. The zero'ing code uses a kernel function called blkdev_issue_zeroout.
Along with the three upstream changes are additional commits to add testing and user mode support. Those commits come directly after the code changes they relate to. The additional user mode support change is for getting blkdev_issue_zeroout to run in user mode.