BTRFS should be able to handle a single broken sector easily. As far as I know, there are multiple levels of checksumming and the ability to repair damages / corruption automatically.
I’m a bit surprised to hear tha COW should increase SSD wearing.
… that should help reduce wearing on SSD - noatime is available for all file systems, but ‘ssd’ is a BTRFS-flag as far as I recall.
As you’re doing a lot of GPU-development, I take it you have a premium PSU, so it’s likely not a PSU problem. If using some cheap PSU found in the trash, power outages from the PSU could cause data corruption; still it shouldn’t cause serious problems if just hitting a few sectors.
It’s impressive that the NTFS kept working on a clicking disk for a month.
Personally, I’m not too fond of NTFS, because we had a server at work, which was set up as RAID1. For some reason, there was data-corruption on one drive, and unfortunately the system decided to replicate the faulty drive onto the healthy one. Such behavior shouldn’t be possible, but I guess the conditions were right in our case. Always keep at least one backup, no matter which FS you’re using.
… well, since harddisks usually have a tolerance of 10% on the input voltage (12V, 5V), then it’s unlikely (but not impossible) that it’s a PSU problem. It sounds to me that there’s plenty of power left in your system. I’ve seen bad drives that came straight from my dealer (new in box), so it could be a faulty drive. Other than that, I’m close to be out of ideas what could have happened (except from evil electrons in the circuits ).
Hello! I have been working on this patch. For testing, I have used btrfs_shell. “touch” command gives “read only”. Upon reviewing the code, there are multiple read only flags set. Although I have bypassed it to some extent, such that inode has been created but overall it is creating mess, as ExtentAllocator seems unable to parse the block groups on the standard btrfs test image.
I wanted to ask, how did the testing happened in 2019. Was it hardcoding all these flags or was it some specific test image?
Regards.
I don’t know. It’s possible that the change wasn’t tested at all, or that the person who submitted it had other local changes that were never sent to Gerrit.
Hello,
Thanks alot for all guidance over past weeks. After doing some deep research in Haiku VM, I have finalized the proposal:
I made sure to follow your suggestions, adopted to patch 1531, added comprehensive review of unimplemented hooks.
Since you are busy these days, I will submit it to portal once it opens, but if you got time to review it once again, I’ll love to hear your recommendations and make changes before final deadline.
Thanks again for guiding me through out the project proposal.
When I tried to open Linux BtrFS partitions in Haiku, it displayed nothing when multiple BtrFS subvolumes/snapshots are present. Is it planned to implement some kind of BtrFS subvolumes support (at least provide access to main/default subvolume)?
Yes. I have planned complete subvolume and snapshot creation in post-GSoC goals. Regarding read access so haiku can atleast locate and mount default subvolume, I’ll add it to my stretch goals. If time runs short, this will be my top post-GSoC priority.
