Arch Linux

tanstaafl

Member

Registered: 2015-11-30

Posts: 2

[SOLVED] Block# in /proc/partitions

From what I've read, the third column in the /proc/partitions file is the number of physical disk blocks in the partition listed on that line. https://www.centos.org/docs/5/html/5.1/ … tions.html

This seems to be a completely different number from the "blocks" listed by the df command or statvfs system function (lengthy examination below*). I am wondering if this is because /proc/partitions is in fact measuring something different than df, or if there's some kind of bug which is causing the wrong number of blocks to be listed in the /proc/partitions file. I'm pretty sure I'm just misunderstanding the number in the file, so if someone could explain what the "#blocks" column means in /proc/partitions and how that differs from the blocks counted by df I'd be grateful!

I apologize if all of this is just meaningless babble.

*For example, looking at the number of blocks in my /tmp partition (/dev/sda2 on my machine), I get this from /proc/partitions:

[nfm@gauss DiskSpaceDisplayer]$ cat /proc/partitions
major minor  #blocks  name

   8       16  117220824 sdb
   8       17       1024 sdb1
   8       18  117218759 sdb2
   8        0  488386584 sda
   8        1    2097152 sda1
   8        2    2097152 sda2
   8        3  484191239 sda3
  11        0    1048575 sr0

However, this is the output of df:

[nfm@gauss ~]$ df "/tmp"
Filesystem     1K-blocks  Used Available Use% Mounted on
/dev/sda2        1998672  3120   1874312   1% /tmp

I'm looking at the difference between the "2097152" listed in one place and "1998672" listed in the other.

If I use the statvfs system function, like so:

            struct statvfs buf;

            statvfs("/tmp", &buf);

            std::cout << data << " " << buf.f_blocks << std::endl;

then I get output of 499668, which matches the df output if I use 4096-byte blocks:

[nfm@gauss ~]$ df -B 4K "/tmp"
Filesystem     4K-blocks  Used Available Use% Mounted on
/dev/sda2         499668   780    468578   1% /tmp

So the results from df and statvfs seem to agree, but they are both different from /proc/partitions; further, there doesn't seem to be a block size I can pass to df that would give me the same size in blocks for "/tmp" as the number in /proc/partitions so my conclusion is they are either measuring something different or there is a bug with /proc/partitions.

Thanks!

Last edited by tanstaafl (2015-11-30 01:43:13)

Trilby

Inspector Parrot

Registered: 2011-11-29

Posts: 30,330

Website

Re: [SOLVED] Block# in /proc/partitions

One is measuring the partition size, the other is measuring the filesystem.  Filesystems reserve some partition space.  It looks like your filesystem size is 5% less than the partition size.  This is a common value I've heard for the space reserved by ext3/ext4.

---

"UNIX is simple and coherent" - Dennis Ritchie; "GNU's Not Unix" - Richard Stallman

tanstaafl

Member

Registered: 2015-11-30

Posts: 2

Re: [SOLVED] Block# in /proc/partitions

One is measuring the partition size, the other is measuring the filesystem.  Filesystems reserve some partition space.  It looks like your filesystem size is 5% less than the partition size.  This is a common value I've heard for the space reserved by ext3/ext4.

Ah, that makes sense. Thanks!