Report information
The Basics
Id:
45123
Status:
resolved
Priority:
Medium/Medium
Queue:

People
Owner:
Nobody in particular
Requestors:
Cc:
AdminCc:

BugTracker
Version Fixed:
9.12.0
Version Found:
(no value)
Versions Affected:
(no value)
Versions Planned:
9.12.0
Priority:
P2 Normal
Severity:
S2 Normal
CVSS Score:
(no value)
CVE ID:
(no value)
Component:
BIND Common
Area:
libisc

Dates
Created:Sat, 22 Apr 2017 02:15:38 -0400
Updated:Mon, 09 Oct 2017 12:56:22 -0400
Closed:Mon, 09 Oct 2017 12:56:21 -0400



This bug tracker is no longer active.

Please go to our Gitlab to submit issues (both feature requests and bug reports) for active projects maintained by Internet Systems Consortium (ISC).

Due to security and confidentiality requirements, full access is limited to the primary maintainers.

Subject: Add memory context flag for memfill with pattern after allocs/frees
Date: Sat, 22 Apr 2017 11:45:25 +0530
To: bind-suggest@isc.org
From: "Mukund Sivaraman" <muks@isc.org>
Add a memory context flag for filling with a pattern after allocs and frees. Extend named -M to support "fill" argument. Memfill should be disabled by default as it has an impact on performance. Mukund

Message body not shown because it is not plain text.

On Wed Apr 26 05:00:27 2017, marka wrote: > Other tools that take -m options need to be updated. This affects "named -M" not "named -m". AFAIK there are no other tools that take a -M option for memory context flags. However, what I'd like is to keep the fill option turned on by default in every tool except named. I'd quite like a compile-time option to have it be turned on by default in named, as well. Then we could have -M nofill or something to disable it. Having memory fill turned on by default when --enable-developer is in use seems like a good idea.
Just a thought (and a too late thought), but instead of doing a custom code, we can just use AddressSanitizer poisoning and let the ASAN do it's job: https://github.com/google/sanitizers/wiki/AddressSanitizerManualPoisoning This would make the code simpler, more standard and allow us to use ASAN to work on the custom mempool.
> However, what I'd like is to keep the fill option turned on by default in > every tool except named. > > I'd quite like a compile-time option to have it be turned on by default in > named, as well. Then we could have -M nofill or something to disable it. > Having memory fill turned on by default when --enable-developer is in use > seems like a good idea. Implemented these suggestions and pushed to the branch. Please have a look.
Subject: Re: [ISC-Bugs #45123] Add memory context flag for memfill with pattern after allocs/frees
Date: Mon, 9 Oct 2017 11:25:17 +0530
To: "Evan Hunt via RT" <bind9-public@isc.org>
From: "Mukund Sivaraman" <muks@isc.org>
The branch looks good. I've pushed a typo fix. I guess this can't be system tested. Maybe the default memfill can be unit tested in isc_mem_get() to check that it is performed, but it can't be deterministically unit tested to check that the filling is not performed. It would also depend on the compile configuration. Mukund
4768. [func] By default, memory is no longer filled with tag values when it is allocated or freed; this improves performance but makes debugging of certain memory issues more difficult. "named -M fill" turns memory filling back on. (Building "configure --enable-developer", turns memory fill on by default again; it can then be disabled with "named -M nofill".) [RT #45123] 9.12.0