Michael, Thomas,
"Michael Kerrisk (man-pages)" <mtk.manpages@xxxxxxxxx> writes:
> I expect that the small change at Thomas P proposes in this patch is
> correct (i.e., iopl(2) operates per-thread, not per-process). I
> remember that you touch the relevant kernel source file often. Perhaps
> you are able to give a quick Ack?
>>
>> man2/iopl.2 | 6 +++---
>> 1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/man2/iopl.2 b/man2/iopl.2
>> index e5b216a14..329095808 100644
>> --- a/man2/iopl.2
>> +++ b/man2/iopl.2
>> @@ -39,7 +39,7 @@ iopl \- change I/O privilege level
>> .BI "int iopl(int " level );
>> .SH DESCRIPTION
>> .BR iopl ()
>> -changes the I/O privilege level of the calling process,
>> +changes the I/O privilege level of the calling thread,
I'm fine with the s/process/thread/ changes. The permissions are really
per thread.
Though the manpage should mention that a thread inherits the permissions
from the parent, i.e. clone() vs. fork(), exec().
>> as specified by the two least significant bits in
>> .IR level .
>> .PP
>> @@ -50,7 +50,7 @@ Since these X servers require access to all 65536 I/O
>> ports, the
>> call is not sufficient.
>> .PP
>> In addition to granting unrestricted I/O port access, running at a higher
>> -I/O privilege level also allows the process to disable interrupts.
>> +I/O privilege level also allows the thread to disable interrupts.
This paragraph became outdated as of
a24ca9976843 ("x86/iopl: Remove legacy IOPL option")
in v5.5. The kernel no longer allows user space to disable
interrupts. It still grants access to _ALL_ 64k ioports.
Also:
> This call is necessary to allow 8514-compatible X servers to run under
> Linux. Since these X servers require access to all 65536 I/O ports,
> the ioperm(2) call is not sufficient.
is outdated.
ioperm() allows to set all 64k bits, but its significantly slower than
iopl(3) because it needs to copy 8k of data on context switch while
iopl(3) just maps an 'all bits set' static bitmap.
Aside of that only really old x servers rely on iopl(3).
Thanks,
tglx
