https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=224479
--- Comment #32 from [email protected] --- (In reply to Conrad Meyer from comment #31) if tmpfs uses large memory so that files on tmpfs were swapped out, the 1st swapoff_all() can fail due to insufficient memory. vfs_unmountall() will unmount and release such tmpfs memory so that 2nd swapoff_all() may be able to swapoff this time; however, if swap spaces left are still NFS files or mdconfig files, we will still get panic. So, I changed swapoff_all() to release in the reverse order swap spaces are added. System adds swap devices at start (given the penalty of mdconfig/NFS, raw devices are best and most suitable for default entries). When memory and swap space go short, I add md-file or NFS files to avoid immediate issues. So, reverse order in swap_all will take outs manually added entries first to reduce chances of panic. Actually... if all file systems are already unmounted, any processes get swapped-in from swap spaces cannot do much useful anyway; indeed, we may not need 2nd(existing) swapoff_all()... -- You are receiving this mail because: You are the assignee for the bug. _______________________________________________ [email protected] mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-bugs To unsubscribe, send any mail to "[email protected]"
