Re: [patch 069/104] lib/string_helpers.c:string_get_size(): remove redundant prefixes

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Thu, 2015-02-12 at 15:01 -0800, akpm@linux-foundation.org wrote:
> From: Rasmus Villemoes <linux@rasmusvillemoes.dk>
> Subject: lib/string_helpers.c:string_get_size(): remove redundant prefixes
> 
> While 3c9f3681d0b4 "[SCSI] lib: add generic helper to print sizes rounded
> to the correct SI range" says that Z and Y are included in preparation for
> 128 bit computers, they just waste .text currently.  If and when we get
> u128, string_get_size needs updating anyway (and ISO needs to come up with
> four more prefixes).

This is rubbish.  It's nothing to do with 128 bits.  This is to do with
disk sizes linux gets attached to.  The current largest device clusters
are Petabytes ... I think we may have some exabyte ones somewhere in the
Academic community, so it's by no means inconcievable we'll have
Zettabyte ones within a few years.  The SCSI standard, with 4k blocks
supports up to 2^76, which is well into Zettabytes.  We obviously run
off the mmap possibilities a lot sooner, because of the byte offsets,
but that's fixable.  Someone will probably start first by passing blocks
into that interface not bytes, so we'd like it not to be based on
assumptions that think 2^64 is the largest possible value.

> Also there's no need to include and test for the NULL sentinel; once we
> reach "E" size is at most 18.  [The test is also wrong; it should be
> units_str[units][i+1]; if we've reached NULL we're already doomed.]

So fix the bug, don't set us up to run off the end of the array.  And
please consult the community which keeps track of this rather than
trying to get it into Linux without review.

James