lkml.org 
[lkml]   [1996]   [Jun]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: possible SCSI device numbering solution
Date
Followup to:  <Pine.LNX.3.91.960625111607.815B-100000@gytha.demon.co.uk>
By author: Bryn Paul Arnold Jones <bpaj@gytha.demon.co.uk>
In newsgroup: linux.dev.kernel
> True, wasting 48bits of minors could be done, but I think that it will be
> a very long time before we have 16bit's of majors (the way we're using
> them now, it will be a very, very long time).
>

But part of the advantage with having larger majors would be that they
could be sparsely allocated. For example, Foobar Associates is making
a line of telepathic communicator cards. I would like to, as Device
Registrar, to allocate a range of majors in advance to Foobar
Associates. This results in poorer utilization. A rule of thumb is
that the density of a numbering space is inversely proportional to the
logarithm of the size. So N bits can, in reality, hold 2^N/N numbered
gizmos.

That being said, I think 16 bits will be enough for a long time as far
as majors is concerned. Minors, on the other hand, can always use the
space. POSIX.1 does require that dev_t is an arithmetric type, which
means that a 64-bit dev_t would require that Linux permits "long long"
in the offical Linux APIs for 32-bit machines. Since "long long" is a
GCC-ism, it seems to me Linus has been avoiding making it mandatory in
user space. There are a few more issues; a 32-bit dev_t would
maintain the alignment of struct stat, which would make backward
compatibility easier to implement.

-hpa

--
PGP public key available - finger hpa@zytor.com
I don't work for Yggdrasil, but they sponsor the linux.* hierarchy.
"The earth is but one country, and mankind its citizens." -- Bahá'u'lláh
Just Say No to Morden * Save Babylon 5: http://www.babylon5.com/cmp/support/


\
 
 \ /
  Last update: 2005-03-22 13:37    [from the cache]
©2003-2011 Jasper Spaans