On May 21, 2009, at 09:29 PM, Mel Flynn wrote: > On Thursday 21 May 2009 18:53:56 Thomas Backman wrote: >> On May 13, 2009, at 03:19 PM, Thomas Backman wrote: > > <snip electrons> > >>> #11 0xffffffff8123c200 in dtrace_copycheck >>> (uaddr=18446744071581326184, kaddr=Variable "kaddr" is not >>> available. >>> ) at dtrace_isa.c:527 > >> In /sys/cddl/dev/dtrace/amd64/dtrace_isa.c: >> static int >> dtrace_copycheck(uintptr_t uaddr, uintptr_t kaddr, size_t size) >> { >> printf("in dtrace_copycheck(), pre-ASSERT:\n"); >> printf("kaddr = %u, kernelbase = %u, size = %d, kaddr+size = %u\n >> \n", >> (unsigned int)kaddr, (unsigned int)kernelbase, (unsigned >> int)size, (unsigned int)(kaddr+size)); >> ASSERT(kaddr >= kernelbase && kaddr + size >= kaddr); >> >> if (uaddr + size >= kernelbase || uaddr + size < uaddr) { >> ... >> >> So, I added two printf statements. What happens? IT BECOMES STABLE. > > I'm no kernel hacker, but.. if you apply the patch below, does it > still panic? > Make sure to get rid of the printf() you added. > > The theory behind this patch is that kernbase isn't initialized at > the time of > that assert, yet code from printf initializes it. > > --- dtrace_isa.c.orig 2009-05-21 21:18:54.000000000 +0200 > +++ dtrace_isa.c 2009-05-21 21:23:40.000000000 +0200 > _at__at_ -40,7 +40,8 _at__at_ > #include <vm/vm_param.h> > #include <vm/pmap.h> > > -extern uintptr_t kernbase; > +//extern uintptr_t kernbase; > +static uintptr_t kernbase = KERNBASE; > uintptr_t kernelbase = (uintptr_t) &kernbase; > > #define INKERNEL(va) (((vm_offset_t)(va)) >= USRSTACK && \ Hmmmmm. Nope, still panics with your patch, unfortunately. So I reverted to my hack, but that doesn't work anymore, either! I did a full buildworld/buildkernel yesterday, WITHOUT csup'ing before, so the source should have stayed the same. Now I get this: # dtrace -n 'syscall::open:entry { trace(copyinstr(arg0)); }' dtrace: description 'syscall::open:entry ' matched 1 probe CPU ID FUNCTION:NAME 0 38977 open:entry 0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef 0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [...snip...] dtrace: error on enabled probe ID 1 (ID 38977: syscall::open:entry): invalid address (0xffffff803e9afae0) in action #1 at DIF offset 28 dtrace: error on enabled probe ID 1 (ID 38977: syscall::open:entry): invalid address (0xffffff803e9afae0) in action #1 at DIF offset 28 dtrace: error on enabled probe ID 1 (ID 38977: syscall::open:entry): invalid address (0xffffff803e9afae0) in action #1 at DIF offset 28 Same error using opensnoop and/or printing and copying in in :::return, so something happened with the kernel (modules): dtrace: error on enabled probe ID 3 (ID 38978: syscall::open:return): invalid address (0xffffff803e9faae0) in action #10 at DIF offset 28 710400 1970 Jan 1 01:00:00 0 1370 5509120 2 vnstat\0 718047 1970 Jan 1 01:00:00 0 1370 5509120 0 vnstat\0 dtrace: error on enabled probe ID 3 (ID 38978: syscall::open:return): invalid address (0xffffff803e9afae0) in action #10 at DIF offset 28 dtrace: error on enabled probe ID 3 (ID 38978: syscall::open:return): invalid address (0xffffff803e9afae0) in action #10 at DIF offset 28 742667 1970 Jan 1 01:00:00 0 1370 46927872 2 �;�����XĻ vnstat\0 750430 1970 Jan 1 01:00:00 0 1370 46927872 0 �;�����XĻ vnstat\0 (If this appears broken, beyond a few characters, that's because it IS on my screen as well.) The address (0xffffff803e9afae0) changes without restarting dtrace, but it appears fairly constant. ----------- Now, after reinstalling the modules and rebooting (rather than kldunload dtraceall && make install && kldload dtraceall), it works with my ugly hack again. Weird. Since it's all modules, why would it not work to unload, recompile and reload? Regards, ThomasReceived on Fri May 22 2009 - 05:32:01 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:48 UTC