Named anonymous regions: Do we need them?

[Gisle D. <gi...@tr...>]

(Apologies if this has already been addressed).

Some anonymous maps appear in /proc/maps with a name, such as [heap]. 
Currently, the oprofile daemon keeps track of such names, and they get 
used in the sample directories to indicate the type of anonymous 
regions,  e.g. {anon:[heap]}.

Currently, this can cause a crash: Because there is a delay between a 
sample being taken and the reading of /proc/pid/maps, the map can have 
changed from an anonymous map to a named map. If the named map contains 
/'s, we end up with something like {anon:/usr/lib/....}. The directory 
parsing code doesn't seem to like /'s inside the curly brackets, and 
ends up throwing an error from the STL string class (AFAIC).

I'm just wondering what the reason was for putting these names in there 
was? Are there any major objections to removing them?

I believe removing the name will have another benefit: In opd_anon.c, 
all maps of every process with anonymous samples get read from 
/proc/pid/maps. Some of the programs we're dealing with have several 
hundred (mostly named) maps. Couple this with a less-than-perfect 
hashing algorithm, there's a lot of list-walking going on when 
processing anon samples. Storing all maps, not just the anonymous ones, 
clearly does not help.

Any insight would be much appreciated.

Cheers,
Gisle

-- 
Gisle Dankel
Principal Engineer
Transitive
Maybrook House, 40 Blackfriars Street, Manchester M3 2EG
Phone: +44 (0)161 836 2343    Fax: +44 (0)161 836 2399
www.transitive.com
* Eliminate the hardware|software dependency

This E-mail and any attachments hereto are strictly confidential and intended solely for the addressee. If you are not the intended addressee please notify the sender by return and delete the message. You must not disclose, forward or copy this E-mail or attachments to any third party without the prior consent of the sender.