Inspired by some changes made in game of trees, I've
implemented a number of speedups in git9.
First, hashing the chunks during deltification with
murmurhash instead of sha1 speeds up the delta search
significantly.
The stretch function was micro-optimized a bit as well,
since that was taking a large portion of the time when
chunking.
Finally, the full path is not stored. We only care about
grouping files with the same name and path. We don't care
about the ordering. Therefore, only the hash of the path
xored with the hash of the diretory is kept, which saves
a bunch of mallocs and string munging.
This reduces the time spent repacking some test repos
significantly.
9front:
% time git/repack
deltifying 97473 objects: 100%
writing 97473 objects: 100%
indexing 97473 objects: 100%
58.85u 1.39s 61.82r git/repack
% time /sys/src/cmd/git/6.repack
deltifying 97473 objects: 100%
writing 97473 objects: 100%
indexing 97473 objects: 100%
43.86u 1.29s 47.51r /sys/src/cmd/git/6.repack
openbsd:
% time git/repack
deltifying 2092325 objects: 100%
writing 2092325 objects: 100%
indexing 2092325 objects: 100%
1589.48u 45.03s 1729.18r git/repack
% time /sys/src/cmd/git/6.repack
deltifying 2092325 objects: 100%
writing 2092325 objects: 100%
indexing 2092325 objects: 100%
1238.68u 41.49s 1373.15r /sys/src/cmd/git/6.repack
go:
% time git/repack
deltifying 529507 objects: 100%
writing 529507 objects: 100%
indexing 529507 objects: 100%
345.32u 7.71s 369.25r git/repack
% time /sys/src/cmd/git/6.repack
deltifying 529507 objects: 100%
writing 529507 objects: 100%
indexing 529507 objects: 100%
248.07u 4.47s 257.59r /sys/src/cmd/git/6.repack
Checking the range of c|0x60 incorrectly classifies many characters
as alphabetic (digits, control characters 0x01-0x20, and punctuation
characters '!'-':'). This prevents tmparse from parsing dates with
a timezone bounded by those characters (for example, "12:11:56 (PDT)").
Instead, just reuse the isalpha macro provided by ctype.h.
When diffing non-regular files, like /dev/null,
pipes, and similar, diff will generate a temp
file to diff against. This is the right thing
to do, but the temp file leaks into the diff.
This patch retains the original file name all
the way through to diff output.
This code is checking the return of devwalk for
a walk resulting in a clone of an open pipe file. However,
devclone ensures that the chan we are cloning is not
currently open.
This is mostly a copy of port/usbxhci.c with PCIWADDR() replaced
by PADDR() and the pci specific code stripped out.
This could be refactored at a later time.
There is a gpio line for the main hub reset that needs to be
asserted and some power management functions that are currently
done by u-boot (using "usb start" command).
We will do these ourselfs once we have the infrastructure for
it in place.
The getserv function has remained mostly unchanged since
1ed. These checks are for handling string local ports
or lack thereof. This was used by datakit at the time,
where 'ports' were service strings. This made the default
datakit service 'login' when a connector did not give
a specific service. The checks for directory traversal also
seemed specifically to guard against a connector from providing
a malicious service string. With datakit gone all current protocols
use numerics as ports so these checks and defaults become uneeded.
1ed reference:
https://github.com/plan9foundation/plan9/tree/1e-1992-09-21
This is a work in progress port to the mntreform2 laptop.
Working so far:
- mmu (same as raspberry pi 3b+)
- arm generic timer
- gicv3
- uart1
- enet
With access to the uart, one can netboot this kernel in u-boot
using the following commands:
> dhcp
> bootm
devproc allows changing the noteid of another process
which opens a race condition in sysrfork(), when deciding
to inherit the noteid of "up" to the child and calling
pidalloc() later to take the reference, the noteid could
have been changed and the childs noteid could have been
freed already in the process.
this bug can only happen when one writes the /proc/n/noteid
file of a another process than your own that is in the
process of forking.
the noteid changing functionality of devproc seems questinable
and seems to be only used by ape's setpgrid() implementation.
Scrollbar was drawn using a wide dark gutter over a white background
whereas other applications (window, sam, ...) use a thinner scrollbar with an
inverse colorscheme.
This makes the scrollbar more consistent with other 9front applications.
When writing to a large variable in /env, we
spam 'Write error', in spite of the env var
working. This is new behavior, as of commit
b90036a062.
It produces a lot of scary, noisy warnings,
which are probably bogus.
When pushing, git/send would sometimes decide we had all the
objects that we'd need to update the remote, and would try
to pack and send the entire history of the repository. This
is because we only set the 'theirs' ref when we had the object.
If we didn't have the object, we would set a zero hash,
then when deciding if we needed to force, we would think
that we were updating a new branch and send everything,
which would fail to update the remote.
A while ago, qwx noticed that we clobbered the exec
bit when merging files. This is not what we want, so
we changed the operator precedence to avoid merging
dirty files implicitly.
But we do want to merge, because it's convenient for
maintaining permissions. So, instead, we should do a
3 way merge of the exec bit.
This patch does that, as well as reverting the rollback
of that change.
While we're here, we adjust the timestamps correctly
in git/branch.
This requires changes to git/fs, because without an open
handler, lib9p allows opening any file with any mode,
which confuses 'test -x'.
there was a diff that went in a while ago to improve
this, but it got backed out because it encounters a
bug in upstream git -- the spec says that a single
ACK should be sent when not using multi-ack modes,
but they send back multiple ones.
This commit brings back the functionality, and works
around the upstream git bug in two different ways.
First, it skips the packets up until it finds the
start of a pack header.
Second, it deduplicates the want messages, which
is what seems to trigger the duplicate ACKs that
cause us trouble.
This patch makes sam's = cmd output what seems to me a more
useful (plumbable) format: /full/path/to/file:addr , where
addr is line(s) under = and rune(s) under =#.
Avoid calling sdgetdev() for every I/O. Instead,
put the SDunit pointer for #S/sdXX/* files in Chan.aux
and keep a reference to SDev between sdopen()/sdclose().
This avoids having to do the sdindex() lookup and
qlock(),incref(),decref() on every read/write
operation. Removal of SDev's is quite rare and only
can happen with pcmcia ide controllers, and i assume
that for that we can assume thet fileservers having
been exited properly and closed their files before
we attempt to remove a device.
The rest is improving waserror() codepaths, making
sure we release the locks for any of the interface
callbacks (verify/online).
Also get rid of tas() and instead only change the
unit's rawopen flag while holding raw qlock.
The change 775a4bea43
"libsec: various changes to tls"
...
4. simply prf code...
... broke the TLS1.0 prf function, missing the fact
that the prf ouput for sha1 and md5 need to be
xored together.
pci uarts are detected late and usually do not contain
the console= parameter logic.
for these, we can just enable them when devuart is reset,
and replay the boot messages once enabled.
this is usefull as it allows us to use these uarts for
kernel debugging in interrupt context.
MBC3 write switches on a>>13, so the RTC register is 5 (0xA000-0xBFFF).
Mask off upper bits of DH register when updating the timer. Only the
lowest bit is part of the day counter.
Use uint for x in timerforward() so that we don't set negative values
for timer registers if it happens to overflow.
Update timer and then latch rather than the other way around.
Otherwise, timer remains static and will overflow after 512 days.
Although git9 always uses the same commit date and author date, other
implementation do make a distinction. Since commit date is more
representative of the commit graph order, use this as a traversal hint
instead of author date.
If the server only supports the dumb protocol, the first 4 bytes of
response will be the initial part of the hash of the first ref.
The http-protocol documentation says that we should fall back to the
dumb protocol when we don't see a content-type of
application/x-$servicename-advertisement. Check this before
attempting to read a smart git packet.
We now keep track of 3 sets during traversal:
- keep: commits we've reached from head commits
- drop: commits we've reached from tail commits
- skip: ancestors of commits in both 'keep' and 'drop'
Commits in 'keep' and/or 'drop' may be added later to the 'skip' set
if we discover later that they are part of a common subgraph of the
head and tail commits.
From these sets we can calculate the commits we are interested in:
lca commits are those in 'keep' and 'drop', but not in 'skip'.
findtwixt commits are those in 'keep', but not in 'drop' or 'skip'.
The "LCA" commit returned is a common ancestor such that there are no
other common ancestors that can reach that commit. Although there can
be multiple commits that meet this criteria, where one is technically
lower on the commit-graph than the other, these cases only happen in
complex merge arrangements and any choice is likely a decent merge
base.
Repainting is now done in paint() directly. When we find a boundary
commit, we switch our paint color to 'skip'. 'skip' painting does
not stop when it hits another color; we continue until we are left
with only 'skip' commits on the queue.
This fixes several mishandled cases in the current algorithm:
1. If we hit the common subgraph from tail commits first (if the tail
commit was newer than the head commit), we ended up traversing the
entire commit graph. This is because we couldn't distinguish
between 'drop' commits that were part of the common subgraph, and
those that were still looking for it.
2. If we traversed through an initial part of the common subgraph from
head commits before reaching it from tail commits, these commits
were returned from findtwixt even though they were also reachable
from tail commits.
3. In the same case as 2, we might end up choosing an incorrect
commit as the LCA, which is an ancestor of the real LCA.
When loading an acme dump file that contains a window with only one
tag line, there are cases where acme hides that window (i.e. not even
its tag is visible).
The following commands reproduce the issue:
% ed <<EOE
1
i
/tmp
/lib/font/bit/pelm/unicode.8.font
/lib/font/bit/pelm/unicode.8.font
0
f 0 5 175 175 1
5 40 175 1 0 /sys/src/cmd/acme/ Del Snarf Get | Look
f 0 4 330 330 3
4 27 330 1 0 /tmp/ Del Snarf Get | Look
.
,w /tmp/test.dump
Q
EOE
% window -dx 900 -dy 600 'acme -l /tmp/test.dump'
This issue was introduced in commit 47b7dc5ccd.
This avoids ipconfig having to explicitely specify the tag
when we want to set route type, as the tag can be provided
implicitely thru the "tag" command.
This adds a new route "t"-flag that enables network address translation,
replacing the source address (and local port) of a forwarded packet to
one of the outgoing interface.
The state for a translation is kept in a new Translation structure,
which contains two Iphash entries, so it can be inserted into the
per protocol 4-tuple hash table, requiering no extra lookups.
Translations have a low overhead (~200 bytes on amd64),
so we can have many of them. They get reused after 5 minutes
of inactivity or when the per protocol limit of 1000 entries
is reached (then the one with longest inactivity is reused).
The protocol needs to export a "forward" function that is responsible
for modifying the forwarded packet, and then handle translations in
its input function for iphash hits with Iphash.trans != 0.
This patch also fixes a few minor things found during development:
- Include the Iphash in the Conv structure, avoiding estra malloc
- Fix ttl exceeded check (ttl < 1 -> ttl <= 1)
- Router should not reply with ttl exceeded for multicast flows
- Extra checks for icmp advice to avoid protocol confusions.
the bug happens when we did the fast exit thru "done" label,
where we would not make sure that theres space in the buffer
for the NUL terminator.
instead, avoid the fast exit and always do the final
adjbuf() that makes sure we have space for the NUL terminator.
remove the pointless pb checks, they'r wrong (should'v
been bp >= buf+bufsz) and adjbuf() already makes sure this
can never happen.
EDID 1.3 section 5 gives a table describing the priority order of
timing information. Use this ordering when constructing the EDID
mode list.
Since aux/vga selects the first mode in the modelist that matches
the given size, it will now select the mode of that size with the
highest preference. Or, if you set vgasize=auto (or some other
string without an 'x'), aux/vga will select the Preferred Detailed
Timing.
This should make it unnecessary to modify vgadb in many cases.
