Watch Out For BCache Corruption Issues On Linux 5.0 & GCC 9

[oiaohm]

ext4 has no fragmentation? Looooooool. I used to have a MySQL production server where MySQL database files occupied over 40K (!) fragments. It was over 10 years ago, so no SSD, and those files read speed was below 10MB/sec. Also, tell me, how would you defrag free space in ext4? I'm not saying NTFS is perfect - in regard to files fragmentation it's an extremely bad filesystem but ext4 is far from excellent: no free space defragmentation, the number of inodes can't be changed after FS creation, no files compression, random files can't be defragmented for no reasons.

no free space defragmentation
Defrag free space on ext4 is total stupid but doable. Yes this is a tooling issue not a file system limitation.

https://manpages.ubuntu.com/manpages/precise/en/man8/resize2fs.8.html

-M Shrink the filesystem to the minimum size and then restore. This has to be done with the file system offline. The code to perform free space defragmentation is in the resize2fs it would be good if you could do a Shrink File system to minimum size without resizing the file structures. So this is not ext4 limitation its the tools are horrible to-do it. the number of inodes can't be changed after FS creation
Not exactly true this is another thing resize2fs can increase or reduce inode count while file system is offline while changing the file systems size. The size of a inode cannot be changed so if you don't change the size of the file system you cannot change the inode count.

files can't be defragmented for no reasons
https://manpages.ubuntu.com/manpages...4defrag.8.html -v will tell you why a file has refused to defrag. In fact at times it can tell you other files that need to go first.

The resize2fs I will give you is a really stupid item to have to use to change inode counts and defrag free space.

[Rallos Zek]

Linux operating systems have problems in HDD/SSD management since 10/15 years at least. HDDs are struggling on I/O operations. The same hard drive used in XP is not so strained making the operation very quick without any noise compared to every linux operating system. This kind of thing happen when the hard drive is too much fragmented or there are problems on the mapping of data.

Really? I always found Linux disk performance to be much better then any version of Windows most noticeably under heavy loads.

[Weasel]

no free space defragmentation
Defrag free space on ext4 is total stupid but doable. Yes this is a tooling issue not a file system limitation.
https://manpages.ubuntu.com/manpages...size2fs.8.html
-M Shrink the filesystem to the minimum size and then restore. This has to be done with the file system offline. The code to perform free space defragmentation is in the resize2fs it would be good if you could do a Shrink File system to minimum size without resizing the file structures. So this is not ext4 limitation its the tools are horrible to-do it. the number of inodes can't be changed after FS creation
Not exactly true this is another thing resize2fs can increase or reduce inode count while file system is offline while changing the file systems size. The size of a inode cannot be changed so if you don't change the size of the file system you cannot change the inode count.

files can't be defragmented for no reasons
https://manpages.ubuntu.com/manpages...4defrag.8.html -v will tell you why a file has refused to defrag. In fact at times it can tell you other files that need to go first.

The resize2fs I will give you is a really stupid item to have to use to change inode counts and defrag free space.

That's like saying just copy all your files somewhere else and re-create the ext filesystem and copy them back. Same kind of stupid "solutions". If it has to be taken offline then it doesn't count.

[ypnos]

That's like saying just copy all your files somewhere else and re-create the ext filesystem and copy them back. Same kind of stupid "solutions". If it has to be taken offline then it doesn't count.

Yes, there is no good solution to practically non-existent problems. The first sentence in the quote is "Defrag free space on ext4 is total stupid". Guess what, it is. Now about the number if inodes, yes this could be wrongly tuned and therefore need adaptation, I wonder how many % of ext users tune the inode count (I do), and how many of these then fuck it up.

"I can't defrag my ext filesystem!" is like saying "I can't install McAfee on Linux!". Funny thing is you can, but you needn't and shouldn't.

[birdie]

no free space defragmentation
Defrag free space on ext4 is total stupid but doable. Yes this is a tooling issue not a file system limitation.

https://manpages.ubuntu.com/manpages/precise/en/man8/resize2fs.8.html

-M Shrink the filesystem to the minimum size and then restore. This has to be done with the file system offline. The code to perform free space defragmentation is in the resize2fs it would be good if you could do a Shrink File system to minimum size without resizing the file structures. So this is not ext4 limitation its the tools are horrible to-do it. the number of inodes can't be changed after FS creation
Not exactly true this is another thing resize2fs can increase or reduce inode count while file system is offline while changing the file systems size. The size of a inode cannot be changed so if you don't change the size of the file system you cannot change the inode count.

files can't be defragmented for no reasons
https://manpages.ubuntu.com/manpages...4defrag.8.html -v will tell you why a file has refused to defrag. In fact at times it can tell you other files that need to go first.

The resize2fs I will give you is a really stupid item to have to use to change inode counts and defrag free space.

What bullocks do we have here. LOL. In Windows I go to disk manager and resize all my volumes using GUI without even rebooting.

In Linux:

1) you must learn shell
2) you must learn how to boot your system to text mode (or change runlevels whatever they are called now)
3) you must learn how to re/un/mount your root/other partitions read only (Linux GUI partition applications tools cannot operate on mounted partitions)
4) you must learn partition tables and how to work with them
5) you must learn to identify partitions
6) you must learn how to create backups (resizing partitions is not exactly foolproof)
7) you must not make mistakes (you can easily kill all your data by using fdisk/gdisk/gparted/etc.).

That still leaves you with a filesystem which internal state you can barely assess since most distros lack a tool which shows its blocks/extents/whatever graphically.

This still requires a ton of work which wouldn't be necessary had ext4 supported free space defragmentation.

Windows 10 license can be quite legally obtained for less than $20.

[blacknova]

Linux operating systems have problems in HDD/SSD management since 10/15 years at least. HDDs are struggling on I/O operations. The same hard drive used in XP is not so strained making the operation very quick without any noise compared to every linux operating system. This kind of thing happen when the hard drive is too much fragmented or there are problems on the mapping of data.

Haven't tried Windows Server, but IMHO desktop Windows systems heavily prioritize UI operations over IO and in general have less throughput than Linux. OTOH on Linux can easily bring UI to halt on heavy IO, but give good throughput.

[birdie]

Yes, there is no good solution to practically non-existent problems. The first sentence in the quote is "Defrag free space on ext4 is total stupid". Guess what, it is. Now about the number if inodes, yes this could be wrongly tuned and therefore need adaptation, I wonder how many % of ext users tune the inode count (I do), and how many of these then fuck it up.

"I can't defrag my ext filesystem!" is like saying "I can't install McAfee on Linux!". Funny thing is you can, but you needn't and shouldn't.

You haven't mentioned compression, have you? It saves a ton of disk space and can make file operations a lot faster. Also a feature not required by anyone.

[bepvte]

I have used BCache for a good two years, but on 5.0 and recent kernels it corrupted before mkfs.xfs even finished running. I said something in the IRC but didnt go further thinking it was a problem with my setup. Im glad this gets solved, and I have no hard feelings to the bcache team but I really wish they had automated tests or more checking because this bug should not have gone unnoticed.

[ypnos]

whatever

So the FUD is in full force now. Old arguments rebuked, now we suddenly talk about something completely different. Nice try!

[bepvte]

NTFS is one of the few modern filesystems that need defragmentation, and thats because its designed pretty horribly. Ext4 places file in a roomy place with plenty of room to expand by default, and tries much harder to avoid fragmentation. I dont understand why you are going onto a linux forum when you feel irreversibly linux is bad, it feels pretty counter productive and not like you want debate but just to cause a ruckus.