When qat(4) is loaded from from loader.conf(5) on FreeBSD stable/13-n245073-145ca12d22a running on Atom CPU C2758 it introduces panic during boot. No dump or backtrace available, only recording from IPMI console where these messages precede the panic: qat0: <Intel C2000 QuickAssist PF> mem 0xdf280000-0xdf29ffff,0xdf310000-0xdf313fff irq 16 at device 11.0 on pci0 qat0: disabling second AE panic: timed sleep before timers are working To reproduce please add these to /boot/loader.conf on the system with Atom C2000 CPU: cryptodev_load="YES" qat_load="YES" qat_c2xxxfw_load="YES"
Are you using GENERIC? In particular, is options EARLY_AP_STARTUP defined?
Yes, options EARLY_AP_STARTUP is defined. This kernel config file is used: include GENERIC ident SUPERMICRO options IPSEC_SUPPORT nooptions IPSEC # RACK options TCPHPTS options RATELIMIT # Route options ROUTE_MPATH options FIB_ALGO Some additional modules are also loaded early: zfs_load="YES" ichwd_load="YES" aio_load="YES" accf_dns_load="YES" cc_htcp_load="YES" carp_load="YES" tcpmd5_load="YES" ipsec_load="YES"
To clarify I post complte /boot/loader.conf zfs_load="YES" autoboot_delay="10" ichwd_load="YES" aio_load="YES" accf_dns_load="YES" cc_htcp_load="YES" carp_load="YES" tcpmd5_load="YES" ipsec_load="YES" net.fibs="5" vm.pmap.pti="0" loader_logo="beastie" kern.hz="1000" kern.geom.label.gptid.enable="0" net.inet.ip.fw.default_to_accept="1" boot_multicons="YES" comconsole_speed="115200" console="comconsole,vidconsole"
Created attachment 236740 [details] proposed patch My apologies, this fell of my plate. I wrote a patch which I think should fix the problem, but I don't have access to any hardware for testing it. If you (or any other users) are able to confirm that the problem is gone, I'll commit.