2. if you would kind enough to indulge a rather vague inquiry: do you have any thoughts about how the locking described in your last comment might interact with mod_apparmor, when the apache2 parent profile (with ~hundreds of apache2 subprofiles) is reloaded, while a set of apache worker processes are under significant request load? (I very much appreciated your previous, detailed descriptions of the profile parsing and reloading procedures, but I'm still having a hard time seeing how any of these steps would cause degradation for such a long *duration* -- like, minutes or even tens of minutes -- particularly if almost all of the subprofiles' compiled profiles are determined to be duplicates of the ones currently loaded)
Thank you so much for your responses so far, and I will be sure to check my spam folder from now on!
Hi! I'm sorry to say that I'm just now noticing this comment (thanks for that go to yahoo's over-zealous spam filter).
I have a couple followup questions:
1. any chance this fix (https:/ /git.kernel. org/pub/ scm/linux/ kernel/ git/torvalds/ linux.git/ commit/ ?id=a125bcda2d0 aee1d98b51c167a ca60fb312572aa) could be helpful w/r/t the behavior described in this bug? (I'm guessing not, or there would be some linkage in this bug tracker, but the part about “fix aa_xattrs_match() may sleep while holding a RCU lock” sounded intriguing)
2. if you would kind enough to indulge a rather vague inquiry: do you have any thoughts about how the locking described in your last comment might interact with mod_apparmor, when the apache2 parent profile (with ~hundreds of apache2 subprofiles) is reloaded, while a set of apache worker processes are under significant request load? (I very much appreciated your previous, detailed descriptions of the profile parsing and reloading procedures, but I'm still having a hard time seeing how any of these steps would cause degradation for such a long *duration* -- like, minutes or even tens of minutes -- particularly if almost all of the subprofiles' compiled profiles are determined to be duplicates of the ones currently loaded)
Thank you so much for your responses so far, and I will be sure to check my spam folder from now on!