Le 07/08/2017 à 10:38, Kengo NAKAHARA a écrit :
Hi [email protected],"intrctl affinity" in NetBSD-current causes panic. Here is the backtrace ==================== panic: kernel diagnostic assertion "mutex_owned(&cpu_lock) || !mp_online" failed: file "/disk4/home/k-nakahara/repos/netbsd-src/sys/arch/x86/x86/idt.c", line 122 fatal breakpoint trap in supervisor mode trap type 1 code 0 rip 0xffffffff802249e5 cs 0x8 rflags 0x246 cr2 0x72ca08412d98 ilevel 0 rsp 0xffffe4010e898dc0 curlwp 0xffffe4010e88a0c0 pid 0.22 lowest kstack 0xffffe4010e8952c0 Stopped in pid 0.22 (system) at netbsd:breakpoint+0x5: leave db{1}> bt breakpoint() at netbsd:breakpoint+0x5 vpanic() at netbsd:vpanic+0x140 ch_voltag_convert_in() at netbsd:ch_voltag_convert_in idt_vec_set() at netbsd:idt_vec_set+0x82 intr_activate_xcall() at netbsd:intr_activate_xcall+0x8a xc_thread() at netbsd:xc_thread+0xd4 ==================== [email protected] bisects and finds the following commit causes this panic. On 2017/08/01 3:54, Maxime Villard wrote:Module Name: src Committed By: maxv Date: Mon Jul 31 18:54:40 UTC 2017 Modified Files: src/sys/arch/x86/x86: intr.c Log Message: Use idt_vec_set instead. To generate a diff of this commit: cvs rdiff -u -r1.101 -r1.102 src/sys/arch/x86/x86/intr.c Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files.I think setgate() must be called with kpreempt_disable() and x86_disable_intr(). However, I don't know why setgate() must be called with mutex_onwed(&cpu_lock). Could you tell me why mutex_onwed(&cpu_lock) is required?
setgate already disables preemption on amd64, but not on i386. I guess this needs to be investigated. Otherwise it's fixed, thanks Maxime
