[Bug 1214043] New: Slow startup in Firefox on OpenSUSE, 3 Lock Files Timeout?
https://bugzilla.suse.com/show_bug.cgi?id=1214043 Bug ID: 1214043 Summary: Slow startup in Firefox on OpenSUSE, 3 Lock Files Timeout? Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Firefox Assignee: factory-mozilla@lists.opensuse.org Reporter: zrenfire@gmail.com QA Contact: qa-bugs@suse.de Target Milestone: --- Found By: --- Blocker: --- Created attachment 868663 --> https://bugzilla.suse.com/attachment.cgi?id=868663&action=edit Firefox profiler of 3 proflie lock attempts Is it just me that's experiencing slow Firefox startups in OpenSUSE? I tied Ubuntu and it had fairly quick startups. * https://old.reddit.com/r/openSUSE/comments/15jygg0/slow_firefox_startup/ * https://bugzilla.mozilla.org/show_bug.cgi?id=1847463 My old tumbleweed PC had a shit GPU setup, so I didn't mind the slow startups. But I now have a Ryzen 5 5600G desktop and I'm still getting slow startups. I copied the old ~/.mozilla/ folder to my new PC, and it's been slow to startup as well. I tried using the https://profiler.firefox.com/ and running with: MOZ_PROFILER_STARTUP=1 MOZ_PROFILER_STARTUP_ENTRIES=20000000 firefox and noticed it's spending 10+10+5 seconds stuck polling the lock files. https://imgur.com/a/0ubcvo8 The lock file appears to be a symbolic link pointing to `localhost:+9999` which I didn't even know links could do. * OpenSUSE Tumbleweed * KDE Plasma Wayland * btrfs I tried: * `firefox --safe-mode` * Deleting/moving `~/.mozilla` + `~/.cache/mozilla` * New User * Plasma XOrg * Kubuntu 23.04 (startup was only 2 seconds) ext4 * Ubuntu 22.04 (startup was only 2 seconds) ext4 ----- As you can see in the screenshots of the profiler (https://i.imgur.com/d5IUXj4.png), the network requests don't even start until 25s in (the network thread doesn't even start). It gets stuck in the locking of the profile. * `XREMain::XRE_mainStartup` calls * `nsRemoteService::LockStartup`, `nsProfileLock::LockWithSymlink`, `nsProfileLock::Lock(...)`, `PR_GetHostByName`, `__new_gethostbyname_r` * `__nscd_gethostbyname_r`, `nscd_gethst_r`, `__nscd_open_socket`, `wait_on_socket` * `poll` (Spends 5000ms = 5s) * `_nss_dns_gethostbyname_r`, `__GI__nss_dns_gethostbyname_r`, `gethostbyname3_context`, `__GI___res_context_search`, `__res_context_querydomain`, `__GI___res_context_query`, `__res_context_send`, `send_dg` * `poll` (Spends 10s) * `LockProfile(nslNativeAppSupport, nslFile, nslFile, nslToolkitProfile, nslProfileLock)` * `nsToolkitProfile::Lock`, `nsToolkitProfile::Lock`, `nsToolkitProfileLock::Init`, `nsToolkitProfileLock::Init`, `nsProfileLock::Lock`, `nsProfileLock::LockWithSymlink`, `PR_GetHostByName`, `__new_gethostbyname_r` * `_nss_dns_gethostbyname_r`, `__GI__nss_dns_gethostbyname_r`, `gethostbyname3_context`, `__GI___res_context_search`, `__res_context_querydomain`, `__GI___res_context_query`, `__res_context_send`, `send_dg` * `poll` (Spends 10s) ----- The only things I can think of to be OpenSUSE exclusive is it's Firewall or BTRFS. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c1 Michael Pujos <pujos.michael@gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |pujos.michael@gmail.com --- Comment #1 from Michael Pujos <pujos.michael@gmail.com> --- I do have this issue since I upgraded to snapshot 20230823. It happens with Firefox, Thunderbird and deadbeef. using strace on either of these executables always shows a poll() system call waiting and timeouting after 25 seconds, early on program startup: read(12, "\1\0\0\0\0\0\0\0", 8) = 8 read(12, 0x7ffd67536c60, 8) = -1 EAGAIN (Resource temporarily unavailable) write(12, "\1\0\0\0\0\0\0\0", 8) = 8 futex(0x55b090cfc940, FUTEX_WAKE_PRIVATE, 2147483647) = 0 close(12) = 0 write(8, "\1\0\0\0\0\0\0\0", 8) = 8 write(8, "\1\0\0\0\0\0\0\0", 8) = 8 write(8, "\1\0\0\0\0\0\0\0", 8) = 8 futex(0x55b090bf4da0, FUTEX_WAKE_PRIVATE, 1) = 1 futex(0x55b090bedda8, FUTEX_WAKE_PRIVATE, 1) = 1 eventfd2(0, EFD_CLOEXEC|EFD_NONBLOCK) = 12 futex(0x55b090bedda8, FUTEX_WAKE_PRIVATE, 1) = 1 poll([{fd=12, events=POLLIN}], 1, 25000) = 0 (Timeout) -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c2 --- Comment #2 from Michael Pujos <pujos.michael@gmail.com> --- This is observed with 6.4.11 kernel but reverting to 6.4.9 does not fix it. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c3 --- Comment #3 from Michael Pujos <pujos.michael@gmail.com> --- It seems to affect any program that uses GTK: gedit, meld, ... -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c5 --- Comment #5 from Michael Pujos <pujos.michael@gmail.com> --- Yes non GNOME setup (I'm using i3) but xdg-desktop-portal-gnome is installed. I'm investigating the bug report you linked to. What's curious is that this issue happened out of the blue. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c6 --- Comment #6 from Michael Pujos <pujos.michael@gmail.com> --- Mystery solved. The problem was caused by the xdg-desktop-portal-gnome package being *installed*. Uninstalling it fixed the issue. This Arch forum thread revealed that fix: https://bbs.archlinux.org/viewtopic.php?id=285590 Longer story: before stumbling on the post above, I noticed that xdg-desktop-portal-gnome.service did not start: Aug 25 12:02:39 p72 systemd[6676]: Dependency failed for Portal service (GNOME implementation). Aug 25 12:02:39 p72 systemd[6676]: xdg-desktop-portal-gnome.service: Job xdg-desktop-portal-gnome.service/start failed with result 'dependency'. The cause is that it depends on graphical-session.target [Unit] Description=Portal service (GNOME implementation) After=graphical-session.target Requisite=graphical-session.target PartOf=graphical-session.target And I start my system with startx (launched automatically after login on vconsole, so without a Desktop Manager), thus it is in multi-user.target and xdg-desktop-portal-gnome.service cannot start. The xdg-desktop-portal-gnome package was installed automatically on my system during one the latest snapshot a few days ago, probably as an optional dependency for something else. So the permanent fix on my system (and other weird people not using a Desktop Manager) is to lock package xdg-desktop-portal-gnome so it is never installed. That also explains why very few user had this issue. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c8 --- Comment #8 from Chris Holland <zrenfire@gmail.com> --- ATM Firefox is starting up in 2 sec for me. I've been waiting to see if I can reproduce my Firefox **Profile** lock slow startup for a couple weeks now. I've no idea what changed, but Firefox start's up within 2sec in OpenSUSE now... And I've no clue why. I had been dealing with this issue for weeks. The only thing I did was update Kubuntu... And they only share the /boot/efi partition. I did do a full update in OpenSUSE before testing too, and it seems there was a kernel update... But I restarted before testing! ----- I'm using a 1Tb M.2 Drive with multiple partitions for OpenSUSE KDE, Kubuntu, Ubuntu. OpenSUSE Yast Partitioner: https://i.imgur.com/KQtyE6d.png Kubuntu Partitioner: https://i.imgur.com/bw3Uusz.png ----- Here's a (filtered) log of journalctl, as well as profile.json + Screenshot timestamps from multiple boots: https://gist.github.com/Zren/f8e33ccdb4b9fffd2894ff9ae70270a7 The weird thing is that I can read some `journalctl` logs with the kubuntu user... in OpenSUSE. At some point or another, I think I'll do a clean reformat and reinstall of OpenSUSE without any other partitions. -- You are receiving this mail because: You are the assignee for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1214043 https://bugzilla.suse.com/show_bug.cgi?id=1214043#c9 --- Comment #9 from Chris Holland <zrenfire@gmail.com> --- After an update a week or two ago (early September) this bug came back. I had played around with the firewall too before restarting so I wasn't sure if it was the update or editing the firewall (toggling samba+samba-client on then off). However, I noticed that in the journalctl logs, the "hostname" switches from "localhost" to "localhost.localdomain" when Firefox loaded in 2 seconds. However when it's taking 30 seconds to load, the hostname is "Client-57". ❯ sudo journalctl --since=2023-09-01 -g "purge" -- Boot XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX -- Sep 04 21:42:21 localhost systemd[1]: Starting Purge old kernels... Sep 04 21:42:21 localhost systemd[1]: Started Purge old kernels. Sep 04 21:42:21 localhost zypper[934]: Preparing to purge obsolete kernels... Sep 04 21:42:32 localhost.localdomain systemd[1]: purge-kernels.service: Deactivated successfully. Sep 04 21:42:32 localhost.localdomain systemd[1]: purge-kernels.service: Consumed 4.612s CPU time. -- Boot XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX -- Sep 10 19:50:47 localhost systemd[1]: Purge old kernels was skipped because of an unmet condition check (ConditionPathExists=/boot/do_purge_kernels). -- Boot XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX -- Sep 11 17:21:22 localhost systemd[1]: Purge old kernels was skipped because of an unmet condition check (ConditionPathExists=/boot/do_purge_kernels). -- Boot XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX -- Sep 14 12:29:38 localhost systemd[1]: Starting Purge old kernels... Sep 14 12:29:38 localhost zypper[1016]: Preparing to purge obsolete kernels... Sep 14 12:29:38 localhost systemd[1]: Started Purge old kernels. Sep 14 12:29:48 Client-57 systemd[1]: purge-kernels.service: Deactivated successfully. Sep 14 12:29:48 Client-57 systemd[1]: purge-kernels.service: Consumed 4.651s CPU time. ❯ hostnamectl Static hostname: (unset) Transient hostname: Client-57 Icon name: computer-desktop Chassis: desktop 🖥 Machine ID: [snip] Boot ID: [snip] Operating System: openSUSE Tumbleweed CPE OS Name: cpe:/o:opensuse:tumbleweed:20230915 Kernel: Linux 6.5.3-1-default Architecture: x86-64 Hardware Vendor: Micro-Star International Co., Ltd. Hardware Model: MS-7C95 Firmware Version: H.80 Firmware Date: Tue 2023-07-04 Firmware Age: 2month 2w 1d Since "ping localhost.localdomain" is fast, but "ping Client-57" times out, I think I've found the root cause. I looked at /etc/hosts and saw: 127.0.0.1 localhost # fallback hostname used by NetworkManager 127.0.0.1 localhost.localdomain So I edited /etc/hosts and added the following, Firefox was fast to startup in 2sec. 127.0.0.1 Client-57 Any ideas what's causing the hostname to be "Client-57"? -- You are receiving this mail because: You are the assignee for the bug.
participants (1)
-
bugzilla_noreply＠suse.com