Commit 2ec2f55f authored by ASHISH KUMAR GOYAL's avatar ASHISH KUMAR GOYAL

Meeting minutes till 21st Sep.

parent 8c131110
# MTP Meeting Minutes # MTP Meeting Minutes
### 2020-09-21(Monday)
- [@asgoyal] Page cache stats module not complete yet. Traversing fd_array for open files gives NULL for all indexes(fd no). Although for init process, this is not the same.
- [@puru] Check by traverse the fd_table list instead.
- [@puru] The problem description written describes the solution and approach instead of problem itself.
- Action items for Thrusday meeting:
- [] Try out traversing the fdtable list and complete the page cache stats module.
- [] Come up with secong version of problem description with these components-
- [] what is the context and general problem?
- [] what are possible requirements?
- [] why interesting/challenging?
- [] related work
- [] solution components
- [] implementation steps
### 2020-09-17(Thrusday)
- [@asgoyal] Uploaded the problem description.
- [@asgoyal] Kernel- traversed through the task list. kernel panic issue.
- [@puru] Possibly Check for null pointer dereferences.
- [@puru] Use fd number for searching instead of using filename.
- Print the page cache information for different scenarios like no reads, many reads etc.
### 2020-09-14(Monday) ### 2020-09-14(Monday)
- Export symbol issue while trying to call a function defined in existing kernel module from a lodable module. - Export symbol issue while trying to call a function defined in existing kernel module from a lodable module.
- Need to export symbol from the file which defines it, which would require kernel compilation. - Need to export symbol from the file which defines it, which would require kernel compilation.
...@@ -29,4 +50,4 @@ ...@@ -29,4 +50,4 @@
- Possible Challenges- - Possible Challenges-
- VM images may be modified when client system is off. eg- login from different machine, defragment, dedup(?). - VM images may be modified when client system is off. eg- login from different machine, defragment, dedup(?).
- Similarly other challenges has to be figured out. - Similarly other challenges has to be figured out.
- Start Reading about page caches in LKD. - Start Reading about page caches in LKD.
\ No newline at end of file
...@@ -2,11 +2,11 @@ ...@@ -2,11 +2,11 @@
To make a kernel module which provides support for persisting page-cache objects. This module may bypass the the kernel page-cache management module to support mapping persistent page-cache objects. This may further be utilised by KVM to reduce the VM start time by directly utilising the persistent page caches in memory. To make a kernel module which provides support for persisting page-cache objects. This module may bypass the the kernel page-cache management module to support mapping persistent page-cache objects. This may further be utilised by KVM to reduce the VM start time by directly utilising the persistent page caches in memory.
###### OS components which may be impacted by this- #### OS components which may be impacted by this-
- **VFS layer**: This may meed VFS layer to maintain some versioning information to validate the page cache states. The file system metadata caches could also be persisted. This further needs some mechanisms for fault tolerance. - **VFS layer**: This may meed VFS layer to maintain some versioning information to validate the page cache states. The file system metadata caches could also be persisted. This further needs some mechanisms for fault tolerance.
- **Physical FileSystem**: The physical filesystem may also need some minute changes to help VFS in validating the page-cache states. - **Physical FileSystem**: The physical filesystem may also need some minute changes to help VFS in validating the page-cache states.
- **Page-cache management**: Since, the page-caches are kept persistent, this also means that the faults(if occurs) would also be persistent. Also, half-update issue may arrise. So, there has to be some mechanism for ensuring page caches consistency at a suitable granularity of atomicity. - **Page-cache management**: Since, the page-caches are kept persistent, this also means that the faults(if occurs) would also be persistent. Also, half-update issue may arrise. So, there has to be some mechanism for ensuring page caches consistency at a suitable granularity of atomicity.
- **Virual memory**: The new virtual addresses has to be mapped correctly to the persistent physical page caches. This would further need mechanisms to ensure data security and other stuffs. - **Virtual memory**: The new virtual addresses has to be mapped correctly to the persistent physical page caches. This would further need mechanisms to ensure data security and other stuffs.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment