- Add comments about which spls apply to which data structures.
- Consistently protect the rnd_samples queue (the queue of
unprocessed samples) at splhigh().
- allow MD code to supply cpu_timestamp() and cpu_havetimestamp() for
an optional higher-resolution clock/roulette wheel source.
- Collect more statistics on the pool state (keeping track of where
collected bits are going, in addition to where they came from).
- Add RNDGETPOOLSTAT ioctl to get the additional stats.
- Flush a few unused rndpool calls.
- XXX XXX Cryptographic changes:
- 32-bit rotate is:
((val << rp->rotate) | (val >> (32 - rp->rotate))),
not
(val << rp->rotate) | (val >> rp->rotate)
or
((val << rp->rotate) | (val >> (31 - rp->rotate)))
- Avoid overloading of rp->rotate and double-rotation of data
(which limited pool mixing somewhat; "rotate" never got above 7).
- Be more paranoid (but probably not paranoid enough) about mixing
output back into the pool. This is an improvement, but it needs
revisiting soon.
We should follow the spirit of some of the recommendations in
the Schneier PRNG papers:
http://www.counterpane.com/yarrow-notes.htmlhttp://www.counterpane.com/pseudorandom_number.html
including:
- two (or more) stage operation for better isolation between inputs
and outputs.
- use of keyed one-way functions (probably better even than
invertible keyed functions like 3DES) at key points in the data flow,
so that breaking the PRNG is clearly as hard as breaking the function.
from devices connected to pckbc:
- Do actual sample collection in pckbc.
- Add rndsource_element_t to the slot data.
- Change pckbc_set_inputhandler() to take an additional argument,
the name of the device, which is (eventually) passed into
rnd_attach_source() to identify the source.
- Change callers of pckbc_set_inputhander() appropriately.
If __HAVE_CPU_TIMESTAMP is defined, <machine/rnd.h> exists and defines
two functions:
1) u_int32_t cpu_timestamp(void);
returns the 32 low order bits of a reasonably high frequency counter.
2) int cpu_havetimestamp(void);
returns non-zero if cpu_timestamp() actually works.
The timestamp counter should run at a frequency greater than 1 MHz
(otherwise, microtime would be sufficient); other than that, the exact
frequency and origin of the counter are unspecified.
* replace `flush by line' function with a `flush by page' funtion, which
also takes an argument to indicate that write-back caches need not
validate its backing memory.
* use this function when allocating page table memory to flush the cache
before mapping it into kernel virtual space.
* also use it in pmap_{zero,copy}_page(), so we can safely use non-cacheable
access in there.
and scsipi_device respectively, with size reduction of ncr53c9x_softc.
Specifying NULL instructs the driver to use default adapter and default
device codes. Every target port has ncr53c9x_attach(sc, NULL, NULL) anyway.
on LP64 systems (and probably the SPARC) since the __cmsg_alignbytes()
changes went in.
- Change file descriptor passing to use CMSG_DATA(), not (cm + 1). This
pretty much has to be done in order to make it work properly on LP64,
and considering that it's been broken this long...
- Use CMSG_SPACE() to determine the mbuf length needed for a given
control message, and CMSG_LEN() to stash in the cmsg_len member.
- Use CMSG_SPACE() and CMSG_DATA() properly; don't define structures
to hold messages, which will not have the correct alignment after
the cmsg header.
the listen socket properly. This caused spurious rpcbind(8)
failures.
- Don't define a "credmsg" structure. Instead, dynamically
allocate a buffer of the correct size of the message with
CMSG_SPACE() and free it when we're done. This gets the
size of the buffer right on LP64 platforms, which has padding
for alignment thrown in.
- Check for non-present or truncated cred messages when recvmsg(2)
returns.
nisimura