<div class="gmail_quote">On Thu, Jul 9, 2009 at 5:06 PM, Jakub Jelinek <span dir="ltr"><<a href="mailto:jakub@redhat.com">jakub@redhat.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">On Thu, Jul 09, 2009 at 10:32:01AM -0400, Matthew Miller wrote:<br>
> Apparently there was some fun with prelink breaking everything in rawhide<br>
> recently: <<a href="https://bugzilla.redhat.com/show_bug.cgi?id=509655" target="_blank">https://bugzilla.redhat.com/show_bug.cgi?id=509655</a>>. I didn't<br>
> notice, because like Pete Zaitcev says in the comments, removing prelink is<br>
> one of the first things I do.<br>
><br>
> I see it as adding unnecessary complexity and fragility, and it makes<br>
> forensic verification difficult. Binaries can't be verified without being<br>
> modified, which is far from ideal. And the error about dependencies having<br>
> changed since prelinking is disturbingly frequent.<br>
><br>
> On the other hand, smart people have worked on it. It's very likely that<br>
> those smart people know things I don't. I can't find any good numbers<br>
> anywhere demonstrating the concrete benefits provided by prelink. Is there<br>
> data out there? Pretty charts and graphs would be nice. The only things I've<br>
> been able to find are old and not very impressive:<br>
><br>
> <<a href="http://crast.us/james/articles/prelink.php" target="_blank">http://crast.us/james/articles/prelink.php</a>><br>
> <<a href="http://smackerelofopinion.blogspot.com/2009/06/does-prelinking-speed-up-boot-times.html" target="_blank">http://smackerelofopinion.blogspot.com/2009/06/does-prelinking-speed-up-boot-times.html</a>><br>
> <<a href="http://reallylongword.org/prelink-2/" target="_blank">http://reallylongword.org/prelink-2/</a>><br>
<br>
</div>Say on x86-64 F11 (with /etc/prelink.conf.d/nss-prelink.conf removed):<br>
LD_DEBUG=statistics /usr/lib64/openoffice.org3/program/swriter.bin<br>
11799:<br>
11799: runtime linker statistics:<br>
11799: total startup time in dynamic loader: 18632372 clock cycles<br>
11799: time needed for relocation: 2975643 clock cycles (15.9%)<br>
11799: number of relocations: 0<br>
11799: number of relocations from cache: 1527<br>
11799: number of relative relocations: 0<br>
11799: time needed to load objects: 13190177 clock cycles (70.7%)<br>
LD_USE_LOAD_BIAS=0 LD_DEBUG=statistics /usr/lib64/openoffice.org3/program/swriter.bin<br>
11817:<br>
11817: runtime linker statistics:<br>
11817: total startup time in dynamic loader: 145813250 clock cycles<br>
11817: time needed for relocation: 129375884 clock cycles (88.7%)<br>
11817: number of relocations: 42207<br>
11817: number of relocations from cache: 126013<br>
11817: number of relative relocations: 149959<br>
11817: time needed to load objects: 15248893 clock cycles (10.4%)<br>
The former is how long it takes before swriter.bin starts up when prelinked,<br>
the latter tells it not to use prelink information and do normal relocation.<br>
Try this on any other GUI app that links against dozens of shared libraries.</blockquote><div><br>Perfect. this is the key point for me. If i mantain server only, not desktop client, without X and the like,<br>no problem to disable it. Isn't right ? I have already do this, for security reason due to centralizzated integrity checker, on +300 production RHEL machine.<br>
</div><div><br>Regards <br></div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div><div class="h5"><br>
--<br>
fedora-devel-list mailing list<br>
<a href="mailto:fedora-devel-list@redhat.com">fedora-devel-list@redhat.com</a><br>
<a href="https://www.redhat.com/mailman/listinfo/fedora-devel-list" target="_blank">https://www.redhat.com/mailman/listinfo/fedora-devel-list</a><br>
</div></div></blockquote></div><br>