This ensures that a given area is mapped at most twice, and greatly
reduces the virtual address space usage.
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Here's a patch that I think we can merge right now. There may be
other places that need this, but this at least points out the
three places that read/write working tree files for git
update-index, checkout and diff respectively. That should cover
a lot of it [jc: git-apply uses an entirely different codepath
both for reading and writing].
Some day we can actually implement it. In the meantime, this
points out a place for people to start. We *can* even start with
a really simple "we do CRLF conversion automatically, regardless
of filename" kind of approach, that just look at the data (all
three cases have the _full_ file data already in memory) and
says "ok, this is text, so let's convert to/from DOS format
directly".
THAT somebody can write in ten minutes, and it would already
make git much nicer on a DOS/Windows platform, I suspect.
And it would be totally zero-cost if you just make it a config
option (but please make it dynamic with the _default_ just being
0/1 depending on whether it's UNIX/Windows, just so that UNIX
people can _test_ it easily).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
The new interface allows an application to temporarily hash a
small number of objects and pretend that they are available in
the object store without actually writing them.
Signed-off-by: Junio C Hamano <junkio@cox.net>
If open_packed_git failed it may have been because the packfile
actually exists and is readable, but some sort of verification
did not pass. In this case open_packed_git left pack_fd filled
in, as the file descriptor is valid. We don't want to leak the
file descriptor, nor do we want to allow someone in the future
to use this packed_git.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Matthias Lederhofer identified a race condition where a Git reader
process was able to locate an object in a packed_git index, but
was then preempted while a `git repack -a -d` ran and completed.
By the time the reader was able to seek in the packfile to get the
object data, the packfile no longer existed on disk.
In this particular case the reader process did not attempt to
open the packfile before it was deleted, so it did not already
have the pack_fd field popuplated. With the packfile itself gone,
there was no way for the reader to open it and fetch the data.
I'm fixing the race condition by teaching find_pack_entry to ignore
a packed_git whose packfile is not currently open and which cannot
be opened. If none of the currently known packs can supply the
object, we will return 0 and the caller will decide the object is
not available. If this is the first attempt at finding an object,
the caller will reprepare_packed_git and try again. If it was
the second attempt, the caller will typically return NULL back,
and an error message about a missing object will be reported.
This patch does not address the situation of a reader which is
being starved out by a tight sequence of `git repack -a -d` runs.
In this particular case the reader will try twice, probably fail
both times, and declare the object in question cannot be found.
As it is highly unlikely that a real world `git repack -a -d` can
complete faster than a reader can open a packfile, so I don't think
this is a huge concern.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Because I want to reuse open_packed_git in a context where I don't
want the process to die if the packfile in question is bogus, I'm
changing its behavior to return error("...") rather than die("...")
when it detects something is wrong with the packfile it was given.
Right now we still must die out of use_pack should open_packed_git
fail, as none of use_pack's callers are prepared to handle a failure
from that function.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
After staring at the comment and the associated for loop, I
realized the comment was completely bogus. The section of
code its talking about is trying to avoid duplicate mapping
of the same packfile.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
There is little point in having the linked list insertion code
appearing in install_packed_git, and then again just 30 lines
further down in the same file.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
We used to call find_pack_entry() twice from read_sha1_file() in order
to avoid printing an error message, when the object did not exist. This
is fixed by moving the call to error() to the only place it really
could be called.
Signed-off-by: Peter Eriksen <s022018@student.dtu.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Way back when Junio developed the 64 bit index topic he came up
with a means of changing the .idx file format so that older Git
clients would recognize that they don't understand the file and
refuse to read it, while newer clients could tell the difference
between the old-style and new-style .idx files. Unfortunately
this wasn't recorded anywhere.
This change documents how we might go about changing the .idx
file format by using a special signature in the first four bytes.
Credit (and possible blame) goes completely to Junio for thinking
up this technique.
The change also modifies the error message of the current Git code
so that users get a recommendation to upgrade their Git software
should this version or later encounter a new-style .idx which it
cannot process. We already do this for the .pack files, but since
we usually process the .idx files first its important that these
files are recognized and encourage an upgrade.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Originally I introduced read_or_die for the purpose of reading
the pack header and trailer, and I was too lazy to print proper
error messages.
Linus Torvalds <torvalds@osdl.org>:
> For a read error, at the very least you have to say WHICH FILE
> couldn't be read, because it's usually a matter of some file just
> being too short, not some system-wide problem.
and of course Linus is right. Make it so.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
With the new-and-improved write_in_full() semantics, where a partial write
simply always returns a real error (and always sets 'errno' when that
happens, including for the disk full case), a lot of the callers of
write_in_full() were just unnecessarily complex.
In particular, there's no reason to ever check for a zero length or
return: if the length was zero, we'll return zero, otherwise, if a disk
full resulted in the actual write() system call returning zero the
write_in_full() logic would have correctly turned that into a negative
return value, with 'errno' set to ENOSPC.
I really wish every "write_in_full()" user would just check against "<0"
now, but this fixes the nasty and stupid ones.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Unfortunately, while {read,write}_in_full do take into account
zero-sized reads/writes; their die and whine variants do not.
I have a repository where there are zero-sized files in
the history that was triggering these things.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
We have a number of badly checked write() calls. Often we are
expecting write() to write exactly the size we requested or fail,
this fails to handle interrupts or short writes. Switch to using
the new write_in_full(). Otherwise we at a minimum need to check
for EINTR and EAGAIN, where this is appropriate use xwrite().
Note, the changes to config handling are much larger and handled
in the next patch in the sequence.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
We have a number of badly checked read() calls. Often we are
expecting read() to read exactly the size we requested or fail, this
fails to handle interrupts or short reads. Add a read_in_full()
providing those semantics. Otherwise we at a minimum need to check
for EINTR and EAGAIN, where this is appropriate use xread().
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
I do not have any proof that this matters to any existing
problems I am seeing, but I do not think of any reason not to do
this.
Signed-off-by: Junio C Hamano <junkio@cox.net>
In some cases we did not even bother to check the return value of
mmap() and just assume it worked. This is bad, because if we are
out of virtual address space the kernel returned MAP_FAILED and we
would attempt to dereference that address, segfaulting without any
real error output to the user.
We are replacing all calls to mmap() with xmmap() and moving all
MAP_FAILED checking into that single location. If a mmap call
fails we try to release enough least-recently-used pack windows
to possibly succeed, then retry the mmap() attempt. If we cannot
mmap even after releasing pack memory then we die() as none of our
callers have any reasonable recovery strategy for a failed mmap.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
If we are about to fail because this process has run out of memory we
should first try to automatically control our appetite for address
space by releasing enough least-recently-used pack windows to gain
back enough memory such that we might actually be able to meet the
current allocation request.
This should help users who have fairly large repositories but are
working on systems with relatively small virtual address space.
Many times we see reports on the mailing list of these users running
out of memory during various Git operations. Dynamically decreasing
the amount of pack memory used when the demand for heap memory is
increasing is an intelligent solution to this problem.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Much like the alloc_report() function can be useful to report on
object allocation statistics while debugging the new pack_report()
function can be useful to report on the behavior of the mmap window
code used for packfile access.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
If a command opens a packfile for only temporary access and does not
install the struct packed_git* into the global packed_git list then
we are unable to unmap any inactive windows within that packed_git,
causing the overall process to exceed core.packedGitLimit.
We cannot force the callers to install their temporary packfile
into the packed_git chain as doing so would allow that (possibly
corrupt but currently being verified) temporary packfile to become
part of the local ODB, which may allow it to be considered for
object resolution when it may not actually be a valid packfile.
So to support unmapping the windows of these temporary packfiles we
also scan the windows of the struct packed_git which was supplied
to use_pack(). Since commands only work with one temporary packfile
at a time scanning the one supplied to use_pack() and all packs
installed into packed_git should cover everything available in
memory.
We also have to be careful to not close the file descriptor of
the packed_git which was handed to use_pack() when all of that
packfile's windows have been unmapped, as we are already past the
open call that would open the packfile and need the file descriptor
to be ready for mmap() after unuse_one_window returns.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
If we are unable to mmap the a region of the packfile with the mmap()
system call there may be a good reason why, such as a closed file
descriptor or out of address space. Reporting the system level
error message can help to debug such problems.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This finally turns on the sliding window behavior for packfile data
access by mapping limited size windows and chaining them under the
packed_git->windows list.
We consider a given byte offset to be within the window only if there
would be at least 20 bytes (one hash worth of data) accessible after
the requested offset. This range selection relates to the contract
that use_pack() makes with its callers, allowing them to access
one hash or one object header without needing to call use_pack()
for every byte of data obtained.
In the worst case scenario we will map the same page of data twice
into memory: once at the end of one window and once again at the
start of the next window. This duplicate page mapping will happen
only when an object header or a delta base reference is spanned
over the end of a window and is always limited to just one page of
duplication, as no sane operating system will ever have a page size
smaller than a hash.
I am assuming that the possible wasted page of virtual address
space is going to perform faster than the alternatives, which
would be to copy the object header or ref delta into a temporary
buffer prior to parsing, or to check the window range on every byte
during header parsing. We may decide to revisit this decision in
the future since this is just a gut instinct decision and has not
actually been proven out by experimental testing.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
To support multiple windows per packfile we need to unmap only one
window at a time from that packfile, leaving any other windows in
place and available for reference.
We treat all windows from all packfiles equally; the least recently
used, not-in-use window across all packfiles will always be closed
first.
If we have unmapped all windows in a packfile then we can also close
the packfile's file descriptor as its possible we won't need to map
any window from that file in the near future. This decision about
when to close the pack file descriptor may need to be revisited in
the future after additional testing on several different platforms
can be performed.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
When we parse the object header or the delta base reference we
don't bother to loop over use_pack() calls. The reason we don't
need to bother with calling use_pack for each byte accessed is that
use_pack will always promise us at least 20 bytes (really the hash
size) after the offset. This promise from use_pack simplifies a
lot of code in the header parsing logic, as well as helps out the
zlib library by ensuring there's always some data for it to consume
during an inflate call.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
When multiple mmaps start getting used for all pack file access it
is not possible to get all data associated with a specific object
in one contiguous memory region. This limitation prevents simply
passing a single address and length to SHA1_Update or to inflate.
Instead we need to loop until we have processed all data of interest.
As we loop over the data we are always interested in reusing the same
window 'cursor', as the prior window will no longer be of any use
to us. This allows the use_pack() call to automatically decrement
the use count of the prior window before setting up access for us
to the next window.
Within each loop we need to make use of the available length output
parameter of use_pack() to tell us how many bytes are available in
the current memory region, as we cannot tell otherwise.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Part of the implementation concept of the sliding mmap window for
pack access is to permit multiple windows per pack to be mapped
independently. Since the inuse_cnt is associated with the mmap and
not with the file, this value is in struct pack_window and needs to
be incremented/decremented for each pack_window accessed by any code.
To faciliate that implementation we need to replace all uses of
use_packed_git() and unuse_packed_git() with a different API that
follows struct pack_window objects rather than struct packed_git.
The way this works is when we need to start accessing a pack for
the first time we should setup a new window 'cursor' by declaring
a local and setting it to NULL:
struct pack_windows *w_curs = NULL;
To obtain the memory region which contains a specific section of
the pack file we invoke use_pack(), supplying the address of our
current window cursor:
unsigned int len;
unsigned char *addr = use_pack(p, &w_curs, offset, &len);
the returned address `addr` will be the first byte at `offset`
within the pack file. The optional variable len will also be
updated with the number of bytes remaining following the address.
Multiple calls to use_pack() with the same window cursor will
update the window cursor, moving it from one window to another
when necessary. In this way each window cursor variable maintains
only one struct pack_window inuse at a time.
Finally before exiting the scope which originally declared the window
cursor we must invoke unuse_pack() to unuse the current window (which
may be different from the one that was first obtained from use_pack):
unuse_pack(&w_curs);
This implementation is still not complete with regards to multiple
windows, as only one window per pack file is supported right now.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
To efficiently support mmaping of multiple regions of the same pack
file we want to keep the pack's file descriptor open while we are
actively working with that pack. So we are now keeping that file
descriptor in packed_git.pack_fd and closing it only after we unmap
the last window.
This is going to increase the number of file descriptors that are
in use at once, however that will be bounded by the total number of
pack files present and therefore should not be very high. It is
a small tradeoff which we may need to revisit after some testing
can be done on various repositories and systems.
For code clarity we also want to seperate out the implementation
of how we open a pack file from the implementation which locates
a suitable window (or makes a new one) from the given pack file.
Since this is a rather large delta I'm taking advantage of doing
it now, in a fairly isolated change.
When we open a pack file we need to examine the header and trailer
without having a mmap in place, as we may only need to mmap
the middle section of this particular pack. Consequently the
verification code has been refactored to make use of the new
read_or_die function.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
The idea behind the sliding mmap window pack reader implementation
is to have multiple mmap regions active against the same pack file,
thereby allowing the process to mmap in only the active/hot sections
of the pack and reduce overall virtual address space usage.
To implement this we need to refactor the mmap related data
(pack_base, pack_use_cnt) out of struct packed_git and move them
into a new struct pack_window.
We are refactoring the code to support a single struct pack_window
per packfile, thereby emulating the prior behavior of mmap'ing the
entire pack file.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Rather than hardcoding the maximum number of bytes which can be
mmapped from pack files we should make this value configurable,
allowing the end user to increase or decrease this limit on a
per-repository basis depending on the size of the repository
and the capabilities of their operating system.
In general users should not need to manually tune such a low-level
setting within the core code, but being able to artifically limit
the number of bytes which we can mmap at once from pack files will
make it easier to craft test cases for the new mmap sliding window
implementation.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
The unpack_entry_gently function currently has only two callers:
the delta base resolution in sha1_file.c and the main loop of
pack-check.c. Both of these must change to using unpack_entry
directly when we implement sliding window mmap logic, so I'm doing
it earlier to help break down the change set.
This may cause a slight performance decrease for delta base
resolution as well as for pack-check.c's verify_packfile(), as
the pack use counter will be incremented and decremented for every
object that is unpacked.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
If ever new object types are added for future extensions then better
have current git version report them as "unknown" instead of
"corrupted".
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
We used to try loose objects first with sha1_object_info(), but packed
objects first with read_sha1_file(). Now, prefer packed objects over loose
ones with sha1_object_info(), too.
Usually the old behaviour would pose no problem, but when you tried to fix
a fscked up repository by inserting a known-good pack,
git cat-file $(git cat-file -t <sha1>) <sha1>
could fail, even when
git cat-file blob <sha1>
would _not_ fail. Worse, a repack would fail, too.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Currently the error e.g. when pushing to a read-only repository is quite
confusing, this attempts to clean it up, unifies error reporting between
various object writers and uses error() on couple more places.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Since keeping a pushed pack or exploding it into loose objects
should be a local repository decision this teaches receive-pack
to decide if it should call unpack-objects or index-pack --stdin
--fix-thin based on the setting of receive.unpackLimit and the
number of objects contained in the received pack.
If the number of objects (hdr_entries) in the received pack is
below the value of receive.unpackLimit (which is 5000 by default)
then we unpack-objects as we have in the past.
If the hdr_entries >= receive.unpackLimit then we call index-pack and
ask it to include our pid and hostname in the .keep file to make it
easier to identify why a given pack has been kept in the repository.
Currently this leaves every received pack as a kept pack. We really
don't want that as received packs will tend to be small. Instead we
want to delete the .keep file automatically after all refs have
been updated. That is being left as room for future improvement.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This allows us to pass just the file name of a pack rather than
the complete path when we want pack-objects to consider its
contents as though they were loose objects. This can be helpful
if $GIT_OBJECT_DIRECTORY contains shell metacharacters which make
it cumbersome to pass complete paths safely in a shell script.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This adds "keep-pack" extension to send-pack vs receive pack protocol,
and makes the receiver invoke "index-pack --stdin --fix-thin".
With this, you can ask send-pack not to explode the result into
loose objects on the receiving end.
I've patched has_sha1_file() to re-check for added packs just
like is done in read_sha1_file() for now, but I think the static
"re-prepare" interface for packs was a mistake. Creation of a
new pack inside a process that needs to read objects in them
back ought to be a rare event, so we are better off making the
callers (such as receive-pack that calls "index-pack --stdin
--fix-thin") explicitly call re-prepare. That way we do not
have to penalize ordinary users of read_sha1_file() and
has_sha1_file().
We would need to fix this someday.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Supposing that both the base and result sizes were both full size 64-bit
values, their encoding would occupy only 9.2 bytes each. Therefore
inflating 64 bytes is way overkill. Limit it to 20 bytes instead which
should be plenty enough for a couple years to come.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Move file name generation from write_sha1_file_prepare() to the one
caller that cares and make it a void function.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Most callers of write_sha1_file_prepare() are only interested in the
resulting hash but don't care about the returned file name or the header.
This patch adds a simple wrapper named hash_sha1_file() which does just
that, and converts potential callers.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This is the missing part to git-pack-objects allowing it to reuse delta
data to/from any of the two delta types. It can reuse delta from any
type, and it outputs base offsets when --allow-delta-base-offset is
provided and the base is also included in the pack. Otherwise it
outputs base sha1 references just like it always did.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Junio C Hamano