dm-snapshot for system updates in Android
Subsystems (2)
Hello everyone!
I hope you will appreciate knowing that we are currently evaluating the use of dm-snapshot to implement a mechanism to obtain revertible, space-efficient system upgrades in Android. More specifically, we are using dm-snapshot-persistent to test the updated device after reboot, then issue a merge in case of success, otherwise, destroy the snapshot.
This new update mechanism is still under evaluation, but its development is openly done in AOSP.
At the current stage, we have a prototype we are happy with, both in terms of space consumption overhead (for the COW device) and benchmarking results for read-write and merge operations.
I would be glad if you could provide some feedback on a few points that I don't have completely clear.
-- Interface stability
To obtain an initial, empty COW device as quick as possible, we force to 0 only its first 32 bit (magic field). This solution looks clear from the kernel code, but can we rely on that for all the kernels with SNAPSHOT_DISK_VERSION == 1? Would you appreciate it if a similar statement is added as part of /Documentation, making this solution more stable? Or maybe I can think of adding an initialization flag to the dm-snapshot table to explicitly request the COW initialization within the kernel?
Another issue we are facing is to be able to know in advance what the minimum COW device size would be for a given update to be able to allocate the right size for the COW device in advance. To do so, we rely on the current COW structure that seems to have kept the same stable shape in the last decade, and compute the total COW size by knowing the number of modified chunks. The formula would be something like that:
table_line_bytes = 64 * 2 / 8;
exceptions_per_chunk = chunk_size_bytes / table_line_bytes;
total_cow_size_chunks = 1 + 1 + modified_chunks
+ modified_chunks / exceptions_per_chunk;This formula seems to be valid for all the recent kernels we checked. Again, can we assume it to be valid for all the kernels for which SNAPSHOT_DISK_VERSION == 1?
-- Alignment
Our approach follows the solution proposed by Mikulas [1].
Being the block alignment of file extents automatically managed by the filesystem, using FIEMAP should have no alignment-related performance issue. But in our implementation we hit a misalignment [2] branch which leads to dmwarning messages [3, 4].
I have a limited experience with the block layer and dm, so I'm still struggling in finding the root cause for this, either in user space or kernel space.
But our benchmarks seems to be good, so we were thinking as last option to rate-limit or directly remove that warning from our kernels as a temporary solution, but we prefer to avoid diverging from mainline. Rate-limiting is a solution that would make sense also to be proposed in the list, but completely removing the warning doesn't seem the right thing to do. Maybe we are benchmarking something else? What do you think?
Many thanks for taking the time to read this, feedbacks would be highly appreciated.
Regards.
Alessio
[1] https://www.redhat.com/archives/dm-devel/2018-October/msg00363.html
[2] https://elixir.bootlin.com/linux/v5.3/source/block/blk-settings.c#L540
[3] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L484
[4] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L1558
Mikulas Patocka
maintainerHi
Quoted 18 lines · Alessio Balsini
Hello everyone!
I hope you will appreciate knowing that we are currently evaluating the use of dm-snapshot to implement a mechanism to obtain revertible, space-efficient system upgrades in Android. More specifically, we are using dm-snapshot-persistent to test the updated device after reboot, then issue a merge in case of success, otherwise, destroy the snapshot.
This new update mechanism is still under evaluation, but its development is openly done in AOSP.At the current stage, we have a prototype we are happy with, both in terms of space consumption overhead (for the COW device) and benchmarking results for read-write and merge operations.
I would be glad if you could provide some feedback on a few points that I don't have completely clear.
Quoted 5 lines
-- Interface stability
To obtain an initial, empty COW device as quick as possible, we force to 0 only its first 32 bit (magic field). This solution looks clear from the kernel code, but can we rely on that for all the kernels with
SNAPSHOT_DISK_VERSION== 1?
It will work, but, to be consistent with lvm, I suggest to overwrite the first 4k with zeroes.
Quoted 7 lines
Would you appreciate it if a similar statement is added as part of /Documentation, making this solution more stable? Or maybe I can think of adding an initialization flag to the dm-snapshot table to explicitly request the COW initialization within the kernel?
Another issue we are facing is to be able to know in advance what the minimum COW device size would be for a given update to be able to allocate the right
This is hard to say, it depends on what the user is doing with the phone. When dm-snapshot runs out of space, it invalidates the whole snapshot. You'll have to monitor the snapshot space very carefully and take action before it fills up.
I suggest - run main system on the origin target and attach a snapshot that will be used for backup of the data overwritten in the origin. If the updated system fails, merge the snapshot back into the origin; if the update succeeds, drop the snapshot. If the user writes too much data to the device, it would invalidate the only the snapshot (so he can't revert anymore), but it would not invalidate the origin and the data would not be lost.
Quoted 5 lines
size for the COW device in advance. To do so, we rely on the current COW structure that seems to have kept the same stable shape in the last decade, and compute the total COW size by knowing the number of modified chunks. The formula would be something like that:
Quoted 8 lines
table_line_bytes = 64 * 2 / 8; exceptions_per_chunk = chunk_size_bytes / table_line_bytes; total_cow_size_chunks = 1 + 1 + modified_chunks + modified_chunks / exceptions_per_chunk;This formula seems to be valid for all the recent kernels we checked. Again, can we assume it to be valid for all the kernels for which
SNAPSHOT_DISK_VERSION== 1?
Yes, we don't plan to change it.
Quoted 11 lines
-- Alignment
Our approach follows the solution proposed by Mikulas [1].
Being the block alignment of file extents automatically managed by the filesystem, using FIEMAP should have no alignment-related performance issue. But in our implementation we hit a misalignment [2] branch which leads to dmwarning messages [3, 4].I have a limited experience with the block layer and dm, so I'm still struggling in finding the root cause for this, either in user space or kernel space.
I don't know. What is the block size of the filesystem? Are all mappings aligned to this block size?
Quoted 10 lines
But our benchmarks seems to be good, so we were thinking as last option to rate-limit or directly remove that warning from our kernels as a temporary solution, but we prefer to avoid diverging from mainline. Rate-limiting is a solution that would make sense also to be proposed in the list, but completely removing the warning doesn't seem the right thing to do. Maybe we are benchmarking something else? What do you think?
Many thanks for taking the time to read this, feedbacks would be highly appreciated.
Quoted 7 lines
Regards.
Alessio[1] https://www.redhat.com/archives/dm-devel/2018-October/msg00363.html
[2] https://elixir.bootlin.com/linux/v5.3/source/block/blk-settings.c#L540
[3] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L484
[4] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L1558
Mikulas
Hi Mikulas,
Thank you for your answer and suggestions.
Quoted 36 lines · Mikulas Patocka
Hi
Quoted 18 lines · Alessio Balsini
Hello everyone!
I hope you will appreciate knowing that we are currently evaluating the use of dm-snapshot to implement a mechanism to obtain revertible, space-efficient system upgrades in Android. More specifically, we are using dm-snapshot-persistent to test the updated device after reboot, then issue a merge in case of success, otherwise, destroy the snapshot.
This new update mechanism is still under evaluation, but its development is openly done in AOSP.At the current stage, we have a prototype we are happy with, both in terms of space consumption overhead (for the COW device) and benchmarking results for read-write and merge operations.
I would be glad if you could provide some feedback on a few points that I don't have completely clear.
Quoted 5 lines
-- Interface stability
To obtain an initial, empty COW device as quick as possible, we force to 0 only its first 32 bit (magic field). This solution looks clear from the kernel code, but can we rely on that for all the kernels with
SNAPSHOT_DISK_VERSION== 1?It will work, but, to be consistent with lvm, I suggest to overwrite the first 4k with zeroes.
Quoted 5 lines
Would you appreciate it if a similar statement is added as part of /Documentation, making this solution more stable? Or maybe I can think of adding an initialization flag to the dm-snapshot table to explicitly request the COW initialization within the kernel?
Quoted 7 lines
Quoted 2 lines
Another issue we are facing is to be able to know in advance what the minimum COW device size would be for a given update to be able to allocate the right
This is hard to say, it depends on what the user is doing with the phone. When dm-snapshot runs out of space, it invalidates the whole snapshot. You'll have to monitor the snapshot space very carefully and take action before it fills up.
I forgot to mention that all the partitions we are updating are read-only, and can only be modified by snapshot-merge. This allows us to establish a direct relation between the required COW device size and the operations performed by the update (i.e. the number of chunks that are going to be modified).
Quoted 8 lines
I suggest - run main system on the origin target and attach a snapshot that will be used for backup of the data overwritten in the origin. If the updated system fails, merge the snapshot back into the origin; if the update succeeds, drop the snapshot. If the user writes too much data to the device, it would invalidate the only the snapshot (so he can't revert anymore), but it would not invalidate the origin and the data would not be lost.
This is an approach we evaluated, but the main reason why we decided for the solution of updating the snapshot and then merging it to the base device is that we want to be sure that the update was successful before permanently change to the base device. For example, if for some reason the update is interrupted, it would be more difficult to roll-back or restore the update. Additionally, if the update wants to resize the partitions, this operation could not be done until reboot.
Quoted 25 lines
Quoted 6 lines
size for the COW device in advance. To do so, we rely on the current COW structure that seems to have kept the same stable shape in the last decade, and compute the total COW size by knowing the number of modified chunks. The formula would be something like that:
Quoted 8 lines
table_line_bytes = 64 * 2 / 8; exceptions_per_chunk = chunk_size_bytes / table_line_bytes; total_cow_size_chunks = 1 + 1 + modified_chunks + modified_chunks / exceptions_per_chunk;This formula seems to be valid for all the recent kernels we checked. Again, can we assume it to be valid for all the kernels for which
SNAPSHOT_DISK_VERSION== 1?Yes, we don't plan to change it.
Quoted 8 lines
-- Alignment
Our approach follows the solution proposed by Mikulas [1].
Being the block alignment of file extents automatically managed by the filesystem, using FIEMAP should have no alignment-related performance issue. But in our implementation we hit a misalignment [2] branch which leads to dmwarning messages [3, 4].
Quoted 6 lines
Quoted 3 lines
I have a limited experience with the block layer and dm, so I'm still struggling in finding the root cause for this, either in user space or kernel space.
I don't know. What is the block size of the filesystem? Are all mappings aligned to this block size?
Here follows a just generated warning coming from a Pixel 4 kernel (4.14):
[ 3093.443808] device-mapper: table: 253:16: adding target device dm-15
caused an alignment inconsistency: physical_block_size=4096, logical_block_size=4096, alignment_offset=61440, start=0
Does this contain all the info you asked for?
I started investigating this issue, but since we didn't notice any performance degradation, I prioritized other things. I'll be hopefully able to get back to this warning in the next months.
Please let me know if I can help you with that or if you need additional information.
Quoted 14 lines
Quoted 14 lines
But our benchmarks seems to be good, so we were thinking as last option to rate-limit or directly remove that warning from our kernels as a temporary solution, but we prefer to avoid diverging from mainline. Rate-limiting is a solution that would make sense also to be proposed in the list, but completely removing the warning doesn't seem the right thing to do. Maybe we are benchmarking something else? What do you think?
Many thanks for taking the time to read this, feedbacks would be highly appreciated.
Regards.
Alessio
Quoted 7 lines
Quoted 4 lines
[1] https://www.redhat.com/archives/dm-devel/2018-October/msg00363.html
[2] https://elixir.bootlin.com/linux/v5.3/source/block/blk-settings.c#L540
[3] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L484
[4] https://elixir.bootlin.com/linux/v5.3/source/drivers/md/dm-table.c#L1558Mikulas
Thanks again,
Alessio
Mikulas Patocka
maintainerQuoted 15 lines · Alessio Balsini
Quoted 8 lines
Quoted 8 lines
-- Alignment
Our approach follows the solution proposed by Mikulas [1].
Being the block alignment of file extents automatically managed by the filesystem, using FIEMAP should have no alignment-related performance issue. But in our implementation we hit a misalignment [2] branch which leads to dmwarning messages [3, 4].Quoted 6 lines
Quoted 3 lines
I have a limited experience with the block layer and dm, so I'm still struggling in finding the root cause for this, either in user space or kernel space.
I don't know. What is the block size of the filesystem? Are all mappings aligned to this block size?
Quoted 7 lines
Here follows a just generated warning coming from a Pixel 4 kernel (4.14):
[ 3093.443808] device-mapper: table: 253:16: adding target device dm-15caused an alignment inconsistency: physical_block_size=4096, logical_block_size=4096, alignment_offset=61440, start=0
Does this contain all the info you asked for?
Look at the function blk_stack_limits - it has various checks that make it return -1. Insert some debugging printk's there and find out which check made the function return -1.
Based on this, we can find out which of the limits triggered the error message.
Quoted 5 lines
I started investigating this issue, but since we didn't notice any performance degradation, I prioritized other things. I'll be hopefully able to get back to this warning in the next months.
Please let me know if I can help you with that or if you need additional information.
Mikulas