Translate file name to disk blocks on UFS

UNIX Solaris
1

Is there any way to translate a file name to the underlying file system's disk blocks/sectors/extents on UFS (Solaris OS on Sparc)?

I found several ways to do it on linux file systems like ext2/3/4, using command like hdparm -- fibmap and filefrag.
I also found one equivalent way to get that mapping using a plugin called filestat on UFS, but I don't want to use a program I downloaded without seeing the source code first because I need to use this solution on customer's servers too.
Any other ideas on how to get the underlying disk blocks for a file?

Thanks.

2

Why do you need that ?

If you want something you can check the source code, I guess a dtrace script analyzing a whole file read operation would help. You would need to flush the cache first though.

3

I need it so I can find the disk blocks a file is in and read straight from the disk afterwards.
I was actually hoping for a solution other than trying to get the source code for filestat. Maybe some other Solaris command or API that can give the same result.

4

Why?

Why do you need to do this? What is your plan, why have you made it?

Reading raw disk blocks is seldom desirable outside of things like device drivers and boot loaders.

1 Like
5

Don't forget that disk block order doesn't necessarily match file sector/chunk/block order. So, spotting file positions might become a problem.

6

I have a big endian server that writes files to a LUN that is on a small endian machine. I then need to read those file from a 3rd small endian server. (can't change that). I thought maybe if I write the physical disk blocks the files are in to the small endian server I can then read those disk blocks directly on the 3rd server to avoid having to somehow read the big endian file-system meta data on a small endian server.

7

Reading raw disk blocks is going to be a lot more effort and a lot less reliable than arranging a better way to get at the file, like a NFS share, batch download, etc.

8

How could you even determine which disk blocks to read without reading the big-endian raw disk data?

9

I'm guessing that info would be read on the server then transferred over the network somehow.

But if you can talk to the server over the network, might as well get the file that way.

10

Then why using UFS and not ZFS which is directly readable/writable whatever the server endianness ?