[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Entropy Patch with Linebreaks
|
From: |
Michael Casadevall |
|
Subject: |
Re: Entropy Patch with Linebreaks |
|
Date: |
Sun, 12 Aug 2007 16:44:59 -0400 |
On Aug 12, 2007, at 5:47 AM, Samuel Thibault wrote:
Hi,
Michael Casadevall, le Sat 11 Aug 2007 23:03:37 -0400, a écrit :
--- device/blkio.c 25 Feb 1997 21:28:13 -0000 1.1.1.1
+++ device/blkio.c 12 Aug 2007 06:35:59 -0000
@@ -149,6 +151,10 @@
do {
prev = next;
next = prev->io_next;
+#ifdef MACH_ENTROPY
+ /* Let's grab the cylinder numbers for entropy. */
+ entropy_putdata (prev, sizeof(io_req_t), ENTROPY_HIGH_QUALITY);
+#endif
} while (next != 0 && prev->io_cylinder == next->io_cylinder);
if (next == 0) {
Mmm, haven't you noticed that this function is actually never used? :)
Anyway, it may still be useful since disk drivers are supposed to use
it. However, I don't see why you're putting the whole list into the
entropy pool, should it be just the request that is to added? (i.e.
just
ior).
Eh, can't hurt to have the code there, I think it is used though for
any device drivers used
within mach itself.
Index: device/cons.c
===================================================================
RCS file: /sources/hurd/gnumach/device/Attic/cons.c,v
retrieving revision 1.2.4.6
diff -u -r1.2.4.6 cons.c
--- device/cons.c 13 Nov 2006 21:30:36 -0000 1.2.4.6
+++ device/cons.c 12 Aug 2007 06:35:59 -0000
@@ -230,9 +234,19 @@
cngetc()
{
if (cn_tab)
+#if defined(MACH_KERNEL) && defined(MACH_ENTROPY)
+ entropy_putchar (cn_tab->cn_dev, ENTROPY_MEDIUM_QUALITY);
+#endif /* MACH_ENTROPY and MACH_ENTROPY */
Technically speaking, that should be entropy_putshort, since cn_dev
is a
short.
I could rename the function to entropy_putwhatever :-)
+ return ((*cn_tab->cn_getc)(cn_tab->cn_dev, 1));
The characher which was returned really deserves adding to the entropy
pool.
Yeah, that was my intention, I changed the code to reflect this.
+ if (romgetc)
+ {
+#if defined (MACH_KERNEL) && defined(MACH_ENTROPY)
+ entropy_putchar(*romgetc, ENTROPY_MEDIUM_QUALITY);
+#endif /* MACH_KERNEL && MACH_ENTROPY */
That should be an entropy_putptr, shouldn't it? Or you can define a
generic entropy_put macro with typeof() & sizeof() which is able to
infer the size.
I'm not adding the pointer itself. I admit, this is however one of
those times
where having templates would be very nice (although they are generic
evil
in compilers that don't support export)
Index: i386/i386at/kd.c
===================================================================
RCS file: /sources/hurd/gnumach/i386/i386at/Attic/kd.c,v
retrieving revision 1.5.2.13
diff -u -r1.5.2.13 kd.c
--- i386/i386at/kd.c 7 May 2007 22:04:53 -0000 1.5.2.13
+++ i386/i386at/kd.c 12 Aug 2007 06:36:00 -0000
@@ -811,6 +816,16 @@
up = TRUE;
scancode &= ~K_UP;
}
+
+#ifdef MACH_KERNEL
+#ifdef MACH_ENTROPY
+ /* Sune Kirkeby's entropy patch (which was a port of the
+ linux entropy drivers for GNU mach) placed the keyboard
+ entropy source here. I looked at that for an idea of where
+ how to do write this driver. */
+ entropy_putchar(scancode | (up ? 0200 : 0), ENTROPY_LOW_QUALITY);
Why low quality? I know people always use the same commands (ls,
mv...)
but the order in which they type them is to my mind a very good source
of entropy. Also, the _time_ when characters are typed should be the
best source of entropy, you should really take that into account.
I think the data itself is low quality, but the updated code adds the
timestamp
at high quality.
diff -u -r1.3.2.8 kd_mouse.c
--- i386/i386at/kd_mouse.c 13 Nov 2006 21:30:36 -0000 1.3.2.8
+++ i386/i386at/kd_mouse.c 12 Aug 2007 06:36:00 -0000
@@ -677,6 +680,11 @@
moved.mm_deltaX = (char)mousebuf[1] + (char)mousebuf[3];
moved.mm_deltaY = (char)mousebuf[2] + (char)mousebuf[4];
+#ifdef MACH_ENTROPY
+ /* Kick some mouse data to the entropy driver. */
+ entropy_putchar((buttonchanges + moved.mm_deltaX
+ + moved.mm_deltaY), ENTROPY_HIGH_QUALITY);
+#endif
Ok for the high quality, same story for the time.
See above.
Index: linux/dev/glue/misc.c
===================================================================
RCS file: /sources/hurd/gnumach/linux/dev/glue/Attic/misc.c,v
retrieving revision 1.2
diff -u -r1.2 misc.c
--- linux/dev/glue/misc.c 18 Sep 2001 21:14:19 -0000 1.2
+++ linux/dev/glue/misc.c 12 Aug 2007 06:36:00 -0000
@@ -224,6 +228,15 @@
void
add_blkdev_randomness (int major)
{
+#ifdef MACH_ENTROPY
+ /* This is useless for good quality, so we'll only use if it
nothing
+ else is available - The problem is that mach only has 1 block
+ device, floppy (major 3 corresponds to Ctrl C) so this is
useless
+ for entropic sources. If we ever get more block devices the
+ quality should be upped for additional entropy. */
+
+ entropy_putchar(major, ENTROPY_POOR_QUALITY);
+#endif
}
People themselves usually have very few devices actually. But the
point
of add_blkdev_randomness is not to push the number of the device, but
the _time_ when it gets called, usually the i/o completion time, which
is a somewhat good source of entropy.
Changed to add timestamp.
diff -u -r1.1.4.7 net.c
--- linux/dev/glue/net.c 27 Mar 2007 22:47:11 -0000 1.1.4.7
+++ linux/dev/glue/net.c 12 Aug 2007 06:36:00 -0000
@@ -299,6 +304,11 @@
ph->length = (skb->len - sizeof (struct ether_header)
+ sizeof (struct packet_header));
+#ifdef MACH_ENTROPY
+ /* Grab the packet for entropy purposes. */
+ entropy_putdata(ph + 1, skb->len - sizeof(struct ether_header),
ENTROPY_HIGH_QUALITY);
+#endif
+
dev_kfree_skb (skb, FREE_READ);
net_kmsg(kmsg)->sent = FALSE; /* Mark packet as received. */
Same here: data is somewhat good, but what is a lot better is the time
when it arrives.
BTW, I think you should add a notion of security: letting network
packets alter the entropy is fine for general desktop purpose, but not
for security purpose: the random numbers could somehow get fooled by
sending the appropriate network packets...
Also, shouldn't you add sent packets to the entropy pool ?
I'm not too worried about security, the mixing code is designed to
make it
that even if you were to constantly enter the same character over and
over
again, it would generate a different number. Also, the mixing code
uses entropy
from the pool to reseed itself, so even if it was possible to
manipulate the pool that
way, you'd also need to control all the other entropic sources, and
have a super
computer to work out what the input would have to be.
diff -u -r1.2 blk.h
--- linux/dev/include/linux/blk.h 5 Apr 2001 06:39:21 -0000 1.2
+++ linux/dev/include/linux/blk.h 12 Aug 2007 06:36:00 -0000
@@ -90,7 +90,7 @@
#endif /* CONFIG_BLK_DEV_MD */
extern void set_device_ro(kdev_t dev,int flag);
-void add_blkdev_randomness(int major);
+extern void add_blkdev_randomness(int major);
extern int floppy_init(void);
extern void rd_load(void);
Well, it's just the same for the compiler :)
I think I changed this because it tossed an error the other way.
Samuel
Michael=
entropy_patch.diff
Description: Binary data
PGP.sig
Description: This is a digitally signed message part