[Crash-utility] [RFC PATCH 1/2] crash: Show memory overcommit data in dump_kmeminfo()

[Aaron Tomlin]

On Fri 2014-11-21 12:06 -0500, Dave Anderson wrote:
> 
> Aaron,
> 
> I like the addition.  But I've got a few nits regarding the patch.
> 
> Given that "kmem -i" is so commonly-used, I prefer not to introduce something
> that could conceivably generate a command-killing "invalid structure member
> offset" error.  And that could happen in your hugetlb_total_pages() function
> because of the hstate structure dependencies.  (See dump_hstates() for example)
> 
> Also, in dump_kmeminfo() you use MEMBER_OFFSET(), which should be avoided
> because it will quietly return -1 if the structure member doesn't exist, 
> leading to bogus output.  That's the whole purpose behind using OFFSET(),
> or at least storing/validating the MEMBER_OFFSET() return value before
> blindly using it.
> 
> It may seem pedantic, but if history has shown us anything, it's that
> kernel developers cannot stop themselves from changing structure/member
> names.

OK - understood.

> So if you could just adjust those kinds of issues, this patch would be
> a nice addition.

Thanks. I'll work on a v2.

-- 
Aaron Tomlin