Searched hist:81 (Results 1 - 25 of 33) sorted by relevance

12

/src/sys/arch/mips/cavium/dev/
H A Docteon_rnm.c1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

1.16 Tue Mar 21 22:07:29 GMT 2023 riastradh branches: 1.16.6;
octrnm(4): Raise delay on startup.

According to CN50XX-HRM-V0.99E and CN78XX-HM-0.99E:

The entropy is provided by the jitter of 125 of 128 free-running
oscillators XORed into a 128-bit LFSR. The LFSR accumulates entropy
over 81 cycles, after which it is fed into a SHA-1 engine.
[...]
The SHA-1 engine runs once every 81 cycles.
[...]
The hardware produces new 64-bit random number every 81 cycles.

The last sentence means that we only need to wait 81 cycles _between_
consecutive SHA-1 outputs (which isn't relevant anyway because we
reconfigure it into raw mode later), but the first two quotes might
mean that we need to wait 81+81 cycles for the _first_ output to be
produced on boot when running the self-test.

Now, in this case, the self-test is run with the LFSR unhooked, by
clearing the RNM_CTL_STATUS[ENT_EN] bit, so that SHA-1 is computed
from a known input -- this is really just paranoia to make sure that
_some_ functions of the device (which is conjured out of thin air at
a fixed virtual address, with no firmware bindings to guide us)
behave as we expect.

And it's not clear if it really does take 81+81 cycles for the first
SHA-1 output to appear when the LFSR isn't feeding into it anyway.
But experimentally, delay of 81+81 cycles seems to work whereas a
delay of only 81 cycles crashes.

PR kern/57280

XXX pullup-10
XXX pullup-9

/src/sys/arch/powerpc/fpu/
H A Dfpu_sqrt.c1.14 Tue Sep 06 23:14:28 GMT 2022 rin Style sync with other parts of this file, as well as FreeBSD:
https://cgit.freebsd.org/src/commit/sys/powerpc/fpu/fpu_sqrt.c?id=81dd9c5e69a2709ae3317dd383093e1c8a970d9e

NFC since q is initialized to zero just above.
1.13 Tue Sep 06 23:12:42 GMT 2022 rin Fix errors in calculation of intermediate mantissa bits > 95.

NFC since this does not affect the final results; we do not
support any extended-precision formats at the moment (never?).

Taken from FreeBSD:
https://cgit.freebsd.org/src/commit/sys/powerpc/fpu/fpu_sqrt.c?id=81dd9c5e69a2709ae3317dd383093e1c8a970d9e
except for stylistic changes that will be committed soon.
/src/tests/usr.bin/xlint/lint1/
H A Dmsg_117.c1.6 Sun Aug 15 13:02:20 GMT 2021 rillig tests/lint: demonstrate wrong warning about signed '>>'

Seen in libdes/ostr2key.c(81).
/src/sys/arch/arm/xscale/
H A Dpxa2x0reg.h1.5 Fri Sep 24 17:30:22 GMT 2004 nathanw PXA255 has 85 GPIO pins, not 81.
(The first revision of the manual listed the wrong number, although it had
all 85 in the table).
/src/sbin/fsck_msdos/
H A Dboot.c1.13 Mon Mar 19 18:30:40 GMT 2007 gdt branches: 1.13.12; 1.13.14; 1.13.16;
Change mismatch of bytes 11 to 90 to be a warning, not an error, and
print out the values of the bytes that do not match.
Add comment explaining that there is no documented rationale for the
check.

fsck_msdos checks that several bytes are equal in the primary and
backup bootblock. There is no documented rationale. The kernel does
not enforce this when mounting a filesystem. I have a FAT32
filesystem on my disk (mounted as /share) which fails this check, but
is accepted by both Windows XP and NetBSD. My volume differs in bytes
71-81. The primary contains "SHARE " and the backup NUL bytes.

Further, fsck_msdos does not have code to repair the mismatch. As a
result such filesystems can be used not checked.
/src/sys/arch/macppc/dev/
H A Dpbms.c1.12 Mon Dec 20 19:18:24 GMT 2010 phx branches: 1.12.8; 1.12.18;
The driver didn't work, because it expects the whole data packet in a
single interrupt. Reality shows that with current kernels a 81 bytes packet
is split into 3 interrupts with 32 + 32 + 17 bytes. I have added a workaround
to deal with it. The Geyser2 devices with 64 byte packets are unsupported
at the moment. Somebody needs to test it.

New: Added support for the iBook 12-inch trackpad.

/src/usr.bin/calendar/calendars/
H A Dcalendar.music1.5 Fri Jul 12 13:42:06 GMT 2002 briggs Comiskey Park was the host of the Disco Demolition in 1979. Home of the
Chicago White Sox for 81 seasons: 1910-1990.

H A Dcalendar.birthday1.30 Fri Aug 23 11:09:14 GMT 2019 sevan Add Danny Cohen
https://www.nytimes.com/2019/08/16/obituaries/danny-cohen-who-helped-set-the-stage-for-a-digital-era-dies-at-81.html
/src/lib/librumpclient/
H A Drumpclient.c1.21 Fri Jan 28 19:21:28 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
1.21 Fri Jan 28 19:21:28 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
/src/lib/librumpuser/
H A Drumpuser_sp.c1.38 Fri Jan 28 19:21:28 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
1.38 Fri Jan 28 19:21:28 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
/src/sys/rump/include/rump/
H A Drumpuser.h1.65 Fri Jan 28 19:21:29 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
1.65 Fri Jan 28 19:21:29 GMT 2011 pooka Pass the value of getprogname() from the client to the server and
record it in p_comm. This is nice for things like sockstat, since
they now display the client command name:

pain-rustique:43:~> rump.sockstat
USER COMMAND PID FD PROTO LOCAL ADDRESS FOREIGN ADDRESS
root xulrunner- 16 0 tcp 192.168.2.114.65507 204.152.190.12.80
root xulrunner- 16 1 tcp 192.168.2.114.65501 204.152.190.12.80
root xulrunner- 16 2 tcp 192.168.2.114.65500 204.152.190.12.80
root xulrunner- 16 3 tcp 192.168.2.114.65499 204.152.190.12.80
root xulrunner- 16 5 tcp 192.168.2.114.65498 204.152.190.12.80
root xulrunner- 16 6 tcp 192.168.2.114.65497 204.152.190.12.80
root socket 62 0 tcp6 *.http *.*
root socket 62 1 tcp *.http *.*
root socket 63 0 tcp6 *.81 *.*
root socket 63 1 tcp *.81 *.*
/src/sys/ufs/ffs/
H A Dffs_bswap.c1.13 Thu Sep 06 02:16:02 GMT 2001 lukem branches: 1.13.4;
Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
H A Dfs.h1.20 Thu Sep 06 02:16:02 GMT 2001 lukem branches: 1.20.2;
Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
/src/sbin/newfs/
H A Dextern.h1.6 Thu Sep 06 02:16:01 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
H A Dnewfs.81.33 Thu Sep 06 02:16:01 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
H A Dnewfs.c1.47 Thu Sep 06 02:16:01 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
/src/sbin/tunefs/
H A Dtunefs.81.26 Thu Sep 06 02:16:01 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
H A Dtunefs.c1.24 Thu Sep 06 02:16:01 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
/src/sys/dev/ic/
H A Drtl81x9var.h1.14 Sat Dec 25 17:24:01 GMT 2004 jonathan Add a forgotten change to rtl81x9var.hfor split re(4) driver
(sys/dev/ic/rtl8169.c): add `sc_rev' to generic 81x9 struct rtk_softc.
H A Dwdcvar.h1.39 Tue Sep 23 09:19:23 GMT 2003 mycroft Fix more probe delay and/or failure problems:
1) Don't wait for DRQ on an IDENTIFY command -- if it's not set when we see
BSY clear, abort the command and ignore the drive. (Do this by testing
for DRQ in the read/write cases in __wdccommand_intr().)
2) Don't wait for DRQ to deassert when we finish an IDENTIFY (or any other
non-block command that reads data) -- we don't do this for block I/O, and
empirically it doesn't clear on my CF cards at all, causing a pointless 1s
delay.
3) Add comments to some of the delay()s, and add missing ones in wdcreset()
and the WDCC_RECAL in the so-called "pre-ATA" probe.
4) Slightly simplify the reset sequence -- we were doing an extra I/O.
5) Modify the register writability test to make sure that registers are not
overlapped -- this can happen in some weird cases with a missing device 1.
6) Check the error register value after the reset -- if it's not 01h or 81h,
as appropriate (see ATA spec), punt.
Tested with a number of ATA-only, ATAPI-only, mixed ATA-ATAPI, CF, and IDE
disk configurations.

Also remove the SINGLE_DRIVE nonsense again.
/src/sys/dev/acpi/
H A Dacpi_ec.c1.82 Tue Apr 28 10:04:32 GMT 2020 jmcneill kern/55206: acpibat reporting broken by acpi_ec.c r1.81

Assume byte instead of qword alignment of the buffer passed to the EC
space handler.
/src/sys/dev/isa/
H A Dwdc_isa.c1.34 Tue Sep 23 09:19:24 GMT 2003 mycroft Fix more probe delay and/or failure problems:
1) Don't wait for DRQ on an IDENTIFY command -- if it's not set when we see
BSY clear, abort the command and ignore the drive. (Do this by testing
for DRQ in the read/write cases in __wdccommand_intr().)
2) Don't wait for DRQ to deassert when we finish an IDENTIFY (or any other
non-block command that reads data) -- we don't do this for block I/O, and
empirically it doesn't clear on my CF cards at all, causing a pointless 1s
delay.
3) Add comments to some of the delay()s, and add missing ones in wdcreset()
and the WDCC_RECAL in the so-called "pre-ATA" probe.
4) Slightly simplify the reset sequence -- we were doing an extra I/O.
5) Modify the register writability test to make sure that registers are not
overlapped -- this can happen in some weird cases with a missing device 1.
6) Check the error register value after the reset -- if it's not 01h or 81h,
as appropriate (see ATA spec), punt.
Tested with a number of ATA-only, ATAPI-only, mixed ATA-ATAPI, CF, and IDE
disk configurations.

Also remove the SINGLE_DRIVE nonsense again.
/src/usr.sbin/dumpfs/
H A Ddumpfs.c1.30 Thu Sep 06 02:16:02 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
1.30 Thu Sep 06 02:16:02 GMT 2001 lukem Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory

The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.

The original FreeBSD commit messages are attached:

=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.

------

One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.

First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:

1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35

2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50

You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html

Test Results

tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66

"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.

------

Algorithm description

The old dirpref algorithm is described in comments:

/*
* Find a cylinder to place a directory.
*
* The policy implemented by this algorithm is to select from
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/

A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.

What I mean by a big file system ?

1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.

The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.

My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/

My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.

My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.

The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:

int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */

These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.

I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.

Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====

=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.

Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.

Reviewed by: mckusick
=====

=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
/src/sys/kern/
H A Dsys_ptrace_common.c1.57 Sat Jun 29 11:37:17 GMT 2019 maxv Fix bug, don't release the reflock if we didn't take it in the first place.
Looks like there are other locking issues in here.

Reported-by: syzbot+81d2c90809163ab1e13c@syzkaller.appspotmail.com
/src/sys/dev/pcmcia/
H A Dwdc_pcmcia.c1.58 Tue Sep 23 09:19:25 GMT 2003 mycroft Fix more probe delay and/or failure problems:
1) Don't wait for DRQ on an IDENTIFY command -- if it's not set when we see
BSY clear, abort the command and ignore the drive. (Do this by testing
for DRQ in the read/write cases in __wdccommand_intr().)
2) Don't wait for DRQ to deassert when we finish an IDENTIFY (or any other
non-block command that reads data) -- we don't do this for block I/O, and
empirically it doesn't clear on my CF cards at all, causing a pointless 1s
delay.
3) Add comments to some of the delay()s, and add missing ones in wdcreset()
and the WDCC_RECAL in the so-called "pre-ATA" probe.
4) Slightly simplify the reset sequence -- we were doing an extra I/O.
5) Modify the register writability test to make sure that registers are not
overlapped -- this can happen in some weird cases with a missing device 1.
6) Check the error register value after the reset -- if it's not 01h or 81h,
as appropriate (see ATA spec), punt.
Tested with a number of ATA-only, ATAPI-only, mixed ATA-ATAPI, CF, and IDE
disk configurations.

Also remove the SINGLE_DRIVE nonsense again.

Completed in 151 milliseconds

12