Yet another kernel change for crash - Crash-utility - Crash Utility List Archives

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Yet another kernel change for crash

crash version 4.0-2.12 is available

crash causes segmentation fault on...

Badari Pulavarty

Monday, 14 November 2005 Mon, 14 Nov '05

12:44 p.m.

Hi Dave, It looks like "mm_struct _rss" now got changed. Now we have "mm_struct _file_rss" instead. http://www.kernel.org/git/? p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4294621f41a85497019fae64341aa5351a1921b7 Thanks, Badari

Reply

Show replies by date

Dave Anderson

Monday, 14 November Mon, 14 Nov

2:06 p.m.

Badari Pulavarty wrote:

Hi Dave, It looks like "mm_struct _rss" now got changed. Now we have "mm_struct _file_rss" instead.

Hi Badari, This can be handled similarly to the way I suggested re: the kmem_cache issue. In fact, there already is this in place to deal with the original name change from mm_struct.rss to mm_struct._rss: MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_rss"); i.e., itt needs yet another qualifier: MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_file_rss"); Can you check that out? I'm also (maybe incorrectly) presuming that you were looking into fixing the kmem_cache issue as well? Is that true? Dave

http://www.kernel.org/git/? p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4294621f41a85497019fae64341aa5351a1921b7 Thanks, Badari -- Crash-utility mailing list Crash-utility(a)redhat.com https://www.redhat.com/mailman/listinfo/crash-utility

Reply

Badari Pulavarty

3:49 p.m.

On Mon, 2005-11-14 at 15:06 -0500, Dave Anderson wrote:

Badari Pulavarty wrote: > Hi Dave, > > It looks like "mm_struct _rss" now got changed. > Now we have "mm_struct _file_rss" instead. > > Hi Badari, This can be handled similarly to the way I suggested re: the kmem_cache issue. In fact, there already is this in place to deal with the original name change from mm_struct.rss to mm_struct._rss: MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_rss"); i.e., itt needs yet another qualifier: MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_rss"); if (!VALID_MEMBER(mm_struct_rss)) MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "_file_rss");

I already made this change and its working fine. Only concern reading comments is, now total rss = file_rss + anon_rss. I wasn't sure if it was the case earlier or "rss" used to represent the whole thing and now got split up.

Can you check that out? I'm also (maybe incorrectly) presuming that you were looking into fixing the kmem_cache issue as well? Is that true?

Well, I hacked up kmem_cache stuff to make it work. I need to figure out a clean way to make sure we don't break backward compatibility. Ofcourse, I could add VALID_MEMBER() every where but code looks ugly. I was thinking of adding something like .. MEMBER_OFFSET_INIT_STRUCTS(kmem_cache_s_name, "kmem_cache_s", "kmem_cache", "name"); Which checks first structure and if INVALID_MEMBER() uses the second structure. What do you think ? Thanks, Badari

Reply

Dave Anderson

4:05 p.m.

Badari Pulavarty wrote:

On Mon, 2005-11-14 at 15:06 -0500, Dave Anderson wrote: > Badari Pulavarty wrote: > > Hi Dave, > > > > It looks like "mm_struct _rss" now got changed. > > Now we have "mm_struct _file_rss" instead. > > > > > Hi Badari, > > This can be handled similarly to the way I suggested re: the > kmem_cache > issue. In fact, there already is this in place to deal with the > original name > change from mm_struct.rss to mm_struct._rss: > > MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); > if (!VALID_MEMBER(mm_struct_rss)) > MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", > "_rss"); > > i.e., itt needs yet another qualifier: > > MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", "rss"); > if (!VALID_MEMBER(mm_struct_rss)) > MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", > "_rss"); > if (!VALID_MEMBER(mm_struct_rss)) > MEMBER_OFFSET_INIT(mm_struct_rss, "mm_struct", > "_file_rss"); > I already made this change and its working fine. Only concern reading comments is, now total rss = file_rss + anon_rss. I wasn't sure if it was the case earlier or "rss" used to represent the whole thing and now got split up.

The additional anon_rss bookkeeping is already in place get_task_mem_usage(), so the change above should be sufficient.

> Can you check that out? I'm also (maybe incorrectly) presuming > that you were looking into fixing the kmem_cache issue as well? > Is that true? Well, I hacked up kmem_cache stuff to make it work. I need to figure out a clean way to make sure we don't break backward compatibility. Ofcourse, I could add VALID_MEMBER() every where but code looks ugly. I was thinking of adding something like .. MEMBER_OFFSET_INIT_STRUCTS(kmem_cache_s_name, "kmem_cache_s", "kmem_cache", "name"); Which checks first structure and if INVALID_MEMBER() uses the second structure. What do you think ?

In other words, collect all the ugliness in one place? I suppose you could do something like that, but you have to check all of the places the "non-minus-one" kmem_cache_s offsets, sizes, and array lengths get initialized. I don't have a problem with adding VALID_MEMBER() checks in those eleven (?) locations -- it's certainly not the first time it's been done. And doing it that way shouldn't break backwards compatibility, because you're only adding the new code when the member is invalid. Thanks, Dave

Reply

Badari Pulavarty

5:42 p.m.

New subject: crash-4.0.2.11 crash fixes for kmem_cache_s and rss changes

Hi Dave, Here is the patch to fix crash to address "kmem_cache_s" and "_rss" name changes - in linux git tree. Thanks, Badari

Reply

Dave Anderson

Tuesday, 15 November Tue, 15 Nov

2:15 p.m.

New subject: crash-4.0.2.11 crash fixes for kmem_cache_s and rsschanges

Badari Pulavarty wrote:

Hi Dave, Here is the patch to fix crash to address "kmem_cache_s" and "_rss" name changes - in linux git tree. Thanks, Badari

Nice. The only thing I changed was the duplication of the two STRUCT_SIZE_INIT(kmem_cache_s, "kmem_cache_s") calls, and the location of the STRUCT_SIZE_INIT(kmem_cache_s, "kmem_cache") call: --- memory.c.badari 2005-11-15 15:12:46.000000000 -0500 +++ memory.c 2005-11-15 15:12:58.000000000 -0500 @@ -341,6 +341,7 @@ vm_init(void) MEMBER_OFFSET_INIT(kmem_cache_s_array, "kmem_cache_s", "array"); ARRAY_LENGTH_INIT(len, NULL, "kmem_cache_s.array", NULL, 0); } else { + STRUCT_SIZE_INIT(kmem_cache_s, "kmem_cache"); MEMBER_OFFSET_INIT(kmem_cache_s_num, "kmem_cache", "num"); MEMBER_OFFSET_INIT(kmem_cache_s_next, "kmem_cache", "next"); MEMBER_OFFSET_INIT(kmem_cache_s_name, "kmem_cache", "name"); @@ -509,9 +510,6 @@ vm_init(void) STRUCT_SIZE_INIT(free_area_struct, "free_area_struct"); STRUCT_SIZE_INIT(zone, "zone"); STRUCT_SIZE_INIT(zone_struct, "zone_struct"); - STRUCT_SIZE_INIT(kmem_cache_s, "kmem_cache_s"); - if (!VALID_STRUCT(kmem_cache_s)) - STRUCT_SIZE_INIT(kmem_cache_s, "kmem_cache"); STRUCT_SIZE_INIT(kmem_bufctl_t, "kmem_bufctl_t"); STRUCT_SIZE_INIT(swap_info_struct, "swap_info_struct"); STRUCT_SIZE_INIT(mm_struct, "mm_struct"); Once again -- many thanks, Dave

Reply

7093

days inactive

7094

days old

devel@lists.crash-utility.osci.io

Manage subscription

5 comments

2 participants

tags (0)

participants (2)

Badari Pulavarty
Dave Anderson