dtrace discuss - Nov 2006 - accessing symbols without type info

I''ve been trying to do something relatively simple, but seem thwarted
on it:

# dtrace -n ''fbt::calcloadavg:entry { printf("CMS_USER: %d,
CMS_SYSTEM:
%d, cpu_waitrq: %d\n", `cpu0.cpu_acct[0], `cpu0.cpu_acct[1], 
`cpu0.cpu_waitrq);}''

results in:

dtrace: invalid probe specifier fbt::calcloadavg:entry { 
printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
`cpu0.cpu_acct[0], `cpu0.cpu_acct[1], `cpu0.cpu_waitrq);}: in action 
list: no symbolic type information is available for unix`cpu0: No type 
information available for symbol

I''ve tried some casts, both as a pointer, and just a cpu_t:

# dtrace -n ''fbt::calcloadavg:entry { cpu = (cpu_t)`cpu0; 
printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
cpu.cpu_acct[0], cpu.cpu_acct[1], cpu.cpu_waitrq);}''
# dtrace -n ''fbt::calcloadavg:entry { cpup = (cpu_t *)&`cpu0; 
printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
cpup->cpu_acct[0], cpup->cpu_acct[1], cpup->cpu_waitrq);}''

but they all result in the same error.  I see why from the source, as 
cpu0 has no type.

I''ve been working around it mostly using `cpu_list-> instead, but is
there any way to tell dtrace "trust me, it''s a cpu_t"?

Thanks,
John

> I''ve been trying to do something relatively simple, but seem
thwarted on it:
> 
> # dtrace -n ''fbt::calcloadavg:entry { printf("CMS_USER: %d,
CMS_SYSTEM:
> %d, cpu_waitrq: %d\n", `cpu0.cpu_acct[0], `cpu0.cpu_acct[1], 
> `cpu0.cpu_waitrq);}''
> 
> results in:
> 
> dtrace: invalid probe specifier fbt::calcloadavg:entry { 
> printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
> `cpu0.cpu_acct[0], `cpu0.cpu_acct[1], `cpu0.cpu_waitrq);}: in action 
> list: no symbolic type information is available for unix`cpu0: No type 
> information available for symbol
> 
> I''ve tried some casts, both as a pointer, and just a cpu_t:
> 
> # dtrace -n ''fbt::calcloadavg:entry { cpu = (cpu_t)`cpu0; 
> printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
> cpu.cpu_acct[0], cpu.cpu_acct[1], cpu.cpu_waitrq);}''
> # dtrace -n ''fbt::calcloadavg:entry { cpup = (cpu_t *)&`cpu0; 
> printf("CMS_USER: %d, CMS_SYSTEM: %d, cpu_waitrq: %d\n", 
> cpup->cpu_acct[0], cpup->cpu_acct[1],
cpup->cpu_waitrq);}''
> 
> but they all result in the same error.  I see why from the source, as 
> cpu0 has no type.
> 
> I''ve been working around it mostly using `cpu_list-> instead,
but is
> there any way to tell dtrace "trust me, it''s a cpu_t"?
> 
> Thanks,
> John
Unfortunately the casting logic in the compiler has to first obtain the
source type for the thing you''re trying to cast to determine if the
cast
is legal, and we''re failing to even get the type since it''s
declared in
assembly code.  I suppose some override pragma is possible for this, but
it doesn''t exist at the moment.  That said, you should be able to use
the cpu[] array w/ cpu[0] to get what you want.

-Mike

-- 
Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/