Skip to contentSkip to conversation

add Via Nehemiah ("xstore") rng support

PATCH Jeff Garzik 2 participants 2 messages Mar 12, 2003 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 9 lines · H. Peter Anvin

In article <[email protected]> of
linux.dev.kernel, you write:

Quoted 6 lines

diff -Nru a/drivers/char/Kconfig b/drivers/char/Kconfig

drivers/char/Kconfig +0 -0
@@ -710,7 +710,7 @@
 	  If you're not sure, say N.
 
Quoted 10 lines
Quoted 6 lines

config HW_RANDOM

drivers/char/Kconfig +1 -1
-	tristate "Intel/AMD H/W Random Number Generator support"
+	tristate "Intel/AMD/Via H/W Random Number Generator support"
	depends on (X86 || IA64) && PCI
	---help---
	  This driver provides kernel-side support for the Random Number

How about changing this to "HW Random Number Generator"?

easily done.

Quoted 17 lines

drivers/char/Kconfig

Quoted 10 lines
drivers/char/Kconfig +10 -0
+static inline u32 xstore(u32 *addr, u32 edx_in)
+{
+	u32 eax_out;
+
+	asm(".byte 0x0F,0xA7,0xC0 /* xstore %%edi (addr=%0) */"
+		:"=m"(*addr), "=a"(eax_out)
+		:"D"(addr), "d"(edx_in));
+
+	return eax_out;
+}

Note that your "=m" (*addr) is never actually used here -- it doesn't affect the instruction encoding, and it only shows up in a comment. Since gcc will generate an instruction mode here, it will be highly confused.

Quoted 3 lines

I am assuming 0xC0 is a modr/m byte, in which case the most sane interpretation of this instruction would be "xstore %eax"; %edi is presumably implicit since you claim it can take a REP prefix...

and yet strangely the asm code seems to be correct :)

        .p2align 4,,15
        .type   via_data_present,@function
via_data_present:
        pushl   %edi
        xorl    %edi, %edi	# zero edi
        movl    $3, %edx        #  edx_in
        movl    %edi, via_rng_datum     #  zero via_rng_datum
        movl    $via_rng_datum, %edi    #  addr
#APP
        .byte 0x0F,0xA7,0xC0 /* xstore %edi (addr=via_rng_datum) */
#NO_APP
        popl    %edi
        andl    $15, %eax       #  eax_out
        setne   %al
        movzbl  %al, %eax

ret

#2 lore raw
Quoted 6 lines · Jeff Garzik
Quoted 6 lines

Note that your "=m" (*addr) is never actually used here -- it doesn't affect the instruction encoding, and it only shows up in a comment. Since gcc will generate an instruction mode here, it will be highly confused.

Quoted 7 lines
Quoted 3 lines

I am assuming 0xC0 is a modr/m byte, in which case the most sane interpretation of this instruction would be "xstore %eax"; %edi is presumably implicit since you claim it can take a REP prefix...

and yet strangely the asm code seems to be correct :)

Right... I got confused by the comment (the one inside the asm) and the fact that "addr" appeared twice. I think it's a bad comment; it draws attention to the wrong thing.

-hpa