Concerns Raised Over The "New" NTFS Linux Driver That Merged Last Year

[issy]

BTW, ntfs-3g can also be considered a dead project. It's not seen any major fixes/new features for at least a couple of years now.

The ntfs-3g GitHub repository shows 73 releases, one or two releases in recent years. The latest one was eight months ago.

[npwx]

Can't say I'm surprised, I don't believe that they care about OSS at all. Last time I've checked some of their partitioning or backup products (don't remember exactly which one) contained copy of Linux, no license terms, no source code, no nothing. Their parent company also has a long history of GPL violations. So in my opinion they just stopped making enough money on this to justify any more effort on their own so they threw it over the fence.

[linuxgeex]

I've had it twice, stopped using NTFS volumes for that after the accidents. Never bothered to report for a simple reason: it's near impossible to reproduce.

I had a 250GB volume, already had 100GB of data on it. Tried to download a large enough file (~12GB) without preallocating it - the file came off broken, the partition couldn't be mounted on a second mount. Had to reboot into Windows to chkdsk it. It showed dozens of scary messages. That was enough.

You are a braver birdie than I. Once upon a time I used ntfsclone --metadata to preserve the FS structure before making changes and then upon seeing my files were good I deleted the backup, only to find later that the NTFS volume had lost a directory and even Windows recovery tools couldn't find it. Luckily it was just installed application stuff that could be replaced, but after that I stopped trying to write NTFS.

Today, in a pinch, if I absolutely must modify an NTFS file then I losetup it and modify the loop device to make sure no actual filesystem structure is changed beyond its data blocks. When I last used Windows (not since 2006 or so) I was using ext2fsd driver to mount my /home partition under Windows, and configured /UserData to be on that volume as well. That solves the whole interchange issue.

Oh, and for a dual-boot system you can safely mkswap on [/media/windows]/pagefile.sys and swap on it. Also if you want to mount NTFS, even readonly, in a dual-boot system, you should disable hibernation.

[birdie]

You are a braver birdie than I.

I've had swap disabled both in Linux and in Windows for more than a decade now. People who insist on SWAP presence even if you have plenty of RAM (I rock 64GB on my desktop and 16GB on my laptop) don't understand why it's needed. I've never used hibernate either because I've got SSDs all around and I don't want to write gigabytes of data to save a few seconds of boot time. Then I'm not even sure hibernation with an SSD is faster than a cold boot. It might not even be the case.

As for writing to Linux partitions from Windows - I've always avoided that. There are no good/totally safe ext4 FS drivers for Windows whatsoever. ext2fsd was abandoned a few years ago as its developer disappeared without a trace.

[sinepgib]

When I last used Windows (not since 2006 or so) I was using ext2fsd driver to mount my /home partition under Windows, and configured /UserData to be on that volume as well. That solves the whole interchange issue.

I'm annoyed I never thought of doing it that way. Indeed, I never thought of replicating the good ol' separate home but for Windows.

This is more like a kernel maintainer being too emotional/angry. Nothing "dies" in GPL universe. It gets taken over, 1000x better code gets written. Bonus: MS gets another lesson about which development model is right.

Things die in the GPL universe, often no other code gets written and MS has nothing to do with this issue.

BTW I purchased a lot from Paragon guys back in the day, they aren't bad developers but they are Windows kernel hackers. They are an old school German software house. They can't adapt to GPL/FSF model of development. These people have secretaries, tickets, customer departments etc etc.

Just how do you think big open source projects work nowadays? Hint: it's mostly not hobbyists in their basements.

[ilgazcl]

I'm annoyed I never thought of doing it that way. Indeed, I never thought of replicating the good ol' separate home but for Windows.



Things die in the GPL universe, often no other code gets written and MS has nothing to do with this issue.



Just how do you think big open source projects work nowadays? Hint: it's mostly not hobbyists in their basements.

For this exact reason, either Ubuntu (very likely) or SuSE (enterprise one) will step in and takeover the project. As SUSE is based in Germany, more likely.

[Awesomeness]

Sounds like the war may have impacted something. Who knows.

Which war was Paragon involved in last November?

[brucethemoose]

I've had swap disabled both in Linux and in Windows for more than a decade now. People who insist on SWAP presence even if you have plenty of RAM (I rock 64GB on my desktop and 16GB on my laptop) don't understand why it's needed. I've never used hibernate either because I've got SSDs all around and I don't want to write gigabytes of data to save a few seconds of boot time. Then I'm not even sure hibernation with an SSD is faster than a cold boot. It might not even be the case.

As for writing to Linux partitions from Windows - I've always avoided that. There are no good/totally safe ext4 FS drivers for Windows whatsoever. ext2fsd was abandoned a few years ago as its developer disappeared without a trace.

winbtrfs works well.

Samsung was supposedly making a Windows f2fs driver, but that hasn't panned out yet :/

You can use WSL if you are desperate, and the partition is on a different disk.

[photom]

I'm kind of disappointed that the ntfs3 kernel module produces a corrupt filesystem in 10-30 min on my SSDs/classic HDDs - I tried many times now to get closer to where the error originates. (Afterwards booting into Win10 to fix the filesystem with chkdsk, then boot back into linux). I reverted back to NTFS-3G now and there is no more corruption, like it was the last years of heavy use. And I have some very IO heavy use cases (like a few TB sized sqlite databases with a lot of updates).

So contrary to some other reports here, NTFS-3G works stable at my end, but ntfs3 kernel module corrupts the filesystem.

Looking at the mailing list there are a lot of errors reported starting april 22 -> https://lore.kernel.org/ntfs3/

[caligula]

The ntfs-3g GitHub repository shows 73 releases, one or two releases in recent years. The latest one was eight months ago.

It was added to Github only recently. Previously you had to download some archive from Tuxera's page. It was pretty hidden, not accessible via the links on their site. Had to use google.