In low memory environments, "nix-env -qa" failed because the fork to
run the pager hit the kernel's overcommit limits. Using posix_spawn
gets around this. (Actually, you have to use posix_spawn with the
undocumented POSIX_SPAWN_USEVFORK flag, otherwise it just uses
fork/exec...)
This substituter connects to a remote host, runs nix-store --serve
there, and then forwards substituter commands on to the remote host and
sends their results to the calling program. The ssh-substituter-hosts
option can be specified as a list of hosts to try.
This is an initial implementation and, while it works, it has some
limitations:
* Only the first host is used
* There is no caching of query results (all queries are sent to the
remote machine)
* There is no informative output (such as progress bars)
* Some failure modes may cause unhelpful error messages
* There is no concept of trusted-ssh-substituter-hosts
Signed-off-by: Shea Levy <shea@shealevy.com>
In particular "libutil" was always a problem because it collides with
Glibc's libutil. Even if we install into $(libdir)/nix, the linker
sometimes got confused (e.g. if a program links against libstore but
not libutil, then ld would report undefined symbols in libstore
because it was looking at Glibc's libutil).
Note that adding --show-trace prevents functions calls from being
tail-recursive, so an expression that evaluates without --show-trace
may fail with a stack overflow if --show-trace is given.
Previously, if the Nix evaluator gets a stack overflow due to a deep
or infinite recursion in the Nix expression, the user gets an
unhelpful message ("Segmentation fault") that doesn't indicate that
the problem is in the user's code rather than Nix itself. Now it
prints:
error: stack overflow (possible infinite recursion)
This only works on x86_64-linux and i686-linux.
Fixes#35.
The integer constant ‘langVersion’ denotes the current language
version. It gets increased every time a language feature is
added/changed/removed. It's currently 1.
The string constant ‘nixVersion’ contains the current Nix version,
e.g. "1.2pre2980_9de6bc5".
This flag causes paths that do not have a known substitute to be
quietly ignored. This is mostly useful for Charon, allowing it to
speed up deployment by letting a machine use substitutes for all
substitutable paths, instead of uploading them. The latter is
frequently faster, e.g. if the target machine has a fast Internet
connection while the source machine is on a slow ADSL line.
I.e. do what git does. I'm too lazy to keep the builtin help text up
to date :-)
Also add ‘--help’ to various commands that lacked it
(e.g. nix-collect-garbage).
Output names are now appended to resulting GC symlinks, e.g. by
nix-build. For backwards compatibility, if the output is named "out",
nothing is appended. E.g. doing "nix-build -A foo" on a derivation
that produces outputs "out", "bin" and "dev" will produce symlinks
"./result", "./result-bin" and "./result-dev", respectively.
By moving the destructor object to libstore.so, it's also run when
download-using-manifests and nix-prefetch-url exit. This prevents
them from cluttering /nix/var/nix/temproots with stale files.
libstore so that the Perl bindings can use it as well. It's vital
that the Perl bindings use the configuration file, because otherwise
nix-copy-closure will fail with a ‘database locked’ message if the
value of ‘use-sqlite-wal’ is changed from the default.
the contents of any of the given store paths have been modified.
E.g.
$ nix-store --verify-path $(nix-store -qR /var/run/current-system)
path `/nix/store/m2smyiwbxidlprfxfz4rjlvz2c3mg58y-etc' was modified! expected hash `fc87e271c5fdf179b47939b08ad13440493805584b35e3014109d04d8436e7b8', got `20f1a47281b3c0cbe299ce47ad5ca7340b20ab34246426915fce0ee9116483aa'
All paths are checked; the exit code is 1 if any path has been
modified, 0 otherwise.
This should also fix:
nix-instantiate: ./../boost/shared_ptr.hpp:254: T* boost::shared_ptr<T>::operator->() const [with T = nix::StoreAPI]: Assertion `px != 0' failed.
which was caused by hashDerivationModulo() calling the ‘store’
object (during store upgrades) before openStore() assigned it.
hook script proper, and the stdout/stderr of the builder. Only the
latter should be saved in /nix/var/log/nix/drvs.
* Allow the verbosity to be set through an option.
* Added a flag --quiet to lower the verbosity level.
This patch adds the configuration file variable "build-cores" and the
command line argument "--cores". These settings specify the number of
CPU cores to utilize for parallel building within a job, i.e. by passing
an appropriate "-j" flag to GNU Make. The default value is 1, which
means that parallel building is *disabled*. If the number of build cores
is specified as 0 (synonymously: "guess" or "auto"), then the actual
value is supposed to be auto-detected by builders at run-time, i.e by
calling the nproc(1) utility from coreutils.
The environment variable $NIX_BUILD_CORES is available to builders, but
the contents of that variable does *not* influence the hash that goes
into the $out store path, i.e. the number of build cores to be utilized
can be changed at will without requiring any re-builds.
This prevents remote builders from being killed by the
`max-silent-time' inactivity monitor while they are waiting for a
long garbage collection to finish. This happens fairly often in the
Hydra build farm.
NixOS evaluation errors in particular look intimidating and
generally aren't very useful. Ideally the builtins.throw messages
should be self-contained.
disasters involving `rm -rf' on bind mounts. Will try the
definitive fix (per-process mounts, apparently possible via the
CLONE_NEWNS flag in clone()) some other time.
it only does something if $NIX_OTHER_STORES (not really a good
name...) is set.
* Do globbing on the elements of $NIX_OTHER_STORES. E.g. you could
set it to /mnts/*/nix or something.
* Install substituters in libexec/nix/substituters.
bytes have been freed, `--max-links' to stop when the Nix store
directory has fewer than N hard links (the latter being important
for very large Nix stores on filesystems with a 32000 subdirectories
limit).
need any info on substitutable paths, we just call the substituters
(such as download-using-manifests.pl) directly. This means that
it's no longer necessary for nix-pull to register substitutes or for
nix-channel to clear them, which makes those operations much faster
(NIX-95). Also, we don't have to worry about keeping nix-pull
manifests (in /nix/var/nix/manifests) and the database in sync with
each other.
The downside is that there is some overhead in calling an external
program to get the substitutes info. For instance, "nix-env -qas"
takes a bit longer.
Abolishing the substitutes table also makes the logic in
local-store.cc simpler, as we don't need to store info for invalid
paths. On the downside, you cannot do things like "nix-store -qR"
on a substitutable but invalid path (but nobody did that anyway).
* Never catch interrupts (the Interrupted exception).
seconds without producing output on stdout or stderr (NIX-65). This
timeout can be specified using the `--max-silent-time' option or the
`build-max-silent-time' configuration setting. The default is
infinity (0).
* Fix a tricky race condition: if we kill the build user before the
child has done its setuid() to the build user uid, then it won't be
killed, and we'll potentially lock up in pid.wait(). So also send a
conventional kill to the child.
`nix-store --delete'. But unprivileged users are not allowed to
ignore liveness.
* `nix-store --delete --ignore-liveness': ignore the runtime roots as
well.
via the Unix domain socket in /nix/var/nix/daemon.socket. The
server forks a worker process per connection.
* readString(): use the heap, not the stack.
* Some protocol fixes.
* Added `build-users-group', the group under which builds are to be
performed.
* Check that /nix/store has 1775 permission and is owner by the
build-users-group.
containing functions that operate on the Nix store. One
implementation is LocalStore, which operates on the Nix store
directly. The next step, to enable secure multi-user Nix, is to
create a different implementation RemoteStore that talks to a
privileged daemon process that uses LocalStore to perform the actual
operations.
gives a huge speedup in operations that read or write from standard
input/output. (So libstdc++'s I/O isn't that bad, you just have to
call std::ios::sync_with_stdio(false).) For instance, `nix-store
--register-substitutes' went from 1.4 seconds to 0.1 seconds on a
certain input. Another victory for Valgrind.
Nix is properly shut down when it receives those signals. In
particular this ensures that killing the garbage collector doesn't
cause a subsequent database recovery.
root (or setuid root), then builds will be performed under one of
the users listed in the `build-users' configuration variables. This
is to make it impossible to influence build results externally,
allowing locally built derivations to be shared safely between
users (see ASE-2005 paper).
To do: only one builder should be active per build user.
immediately add the result as a permanent GC root. This is the only
way to prevent a race with the garbage collector. For instance, the
old style
ln -s $(nix-store -r $(nix-instantiate foo.nix)) \
/nix/var/nix/gcroots/result
has two time windows in which the garbage collector can interfere
(by GC'ing the derivation and the output, respectively). On the
other hand,
nix-store --add-root /nix/var/nix/gcroots/result -r \
$(nix-instantiate --add-root /nix/var/nix/gcroots/drv \
foo.nix)
is safe.
* nix-build: use `--add-root' to prevent GC races.
`derivations.cc', etc.
* Store the SHA-256 content hash of store paths in the database after
they have been built/added. This is so that we can check whether
the store has been messed with (a la `rpm --verify').
* When registering path validity, verify that the closure property
holds.
* Start cleaning up unique store path generation (they weren't always
unique; in particular the suffix ("-aterm-2.2", "-builder.sh") was
not part of the hash, therefore changes to the suffix would cause
multiple store objects with the same hash).
permission to the Nix store or database. E.g., `nix-env -qa' will
work, but `nix-env -qas' won't (the latter needs DB access). The
option `--readonly-mode' forces this mode; otherwise, it's only
activated when the database cannot be opened.
profile. Arguments are either generation number, or `old' to delete
all non-current generations. Typical use:
$ nix-env --delete-generations old
$ nix-collect-garbage
* istringstream -> string2Int.
Previously there was the problem that all files read by nix-env
etc. should be reachable and readable by the Nix user. So for
instance building a Nix expression in your home directory meant that
the home directory should have at least g+x or o+x permission so
that the Nix user could reach the Nix expression. Now we just
switch back to the original user just prior to reading sources and
the like. The places where this happens are somewhat arbitrary,
however. Any scope that has a live SwitchToOriginalUser object in
it is executed as the original user.
* Back out r1385. setreuid() sets the saved uid to the new
real/effective uid, which prevents us from switching back to the
original uid. setresuid() doesn't have this problem (although the
manpage has a bug: specifying -1 for the saved uid doesn't leave it
unchanged; an explicit value must be specified).
more common than the latter (which exists only on Linux and
FreeBSD). We don't really care about dropping the saved IDs since
there apparently is no way to quiry them in any case, so it can't
influence the build (unlike the effective IDs which are checked by
Perl for instance).
setuid installation, since the calling user may have a more fascist
umask (say, 0077), which would cause the store objects built by Nix
to be unreadable to anyone other than the Nix user.
set the real uid and gid to the effective uid and gid, the Nix
binaries can be installed as owned by the Nix user and group instead
of root, so no root involvement of any kind is necessary.
Linux and FreeBSD have these functions.
users.
If the configure flag `--enable-setuid' is used, the Nix programs
nix-env, nix-store, etc. are installed with the setuid bit turned on
so that they are executed as the user and group specified by
`--with-nix-user=USER' and `--with-nix-group=GROUP', respectively
(with defaults `nix' and `nix').
The setuid programs drop all special privileges if they are executed
by a user who is not a member of the Nix group.
The setuid feature is a quick hack to enable sharing of a Nix
installation between users who trust each other. It is not
generally secure, since any user in the Nix group can modify (by
building an appropriate derivation) any object in the store, and for
instance inject trojans into binaries used by other users.
The setuid programs are owned by root, not the Nix user. This is
because on Unix normal users cannot change the real uid, only the
effective uid. Many programs don't work properly when the real uid
differs from the effective uid. For instance, Perl will turn on
taint mode. However, the setuid programs drop all root privileges
immediately, changing all uids and gids to the Nix user and group.
* Builder output is written to standard error by default.
* The option `-B' is gone.
* The option `-Q' suppresses builder output.
The result of this is that most Nix invocations shouldn't need any
flags w.r.t. logging.
Whenever Nix attempts to realise a derivation for which a closure is
already known, but this closure cannot be realised, fall back on
normalising the derivation.
The most common scenario in which this is useful is when we have
registered substitutes in order to perform binary distribution from,
say, a network repository. If the repository is down, the
realisation of the derivation will fail. When this option is
specified, Nix will build the derivation instead. Thus, binary
installation falls back on a source installation. This option is
not the default since it is generally not desirable for a transient
failure in obtaining the substitutes to lead to a full build from
source (with the related consumption of resources).
much as possible. (This is similar to GNU Make's `-k' flag.)
* Refactoring to implement this: previously we just bombed out when
a build failed, but now we have to clean up. In particular this
means that goals must be freed quickly --- they shouldn't hang
around until the worker exits. So the worker now maintains weak
pointers in order not to prevent garbage collection.
* Documented the `-k' and `-j' flags.
in parallel. Hooks are more efficient: locks on output paths are
only acquired when the hook says that it is willing to accept a
build job. Hooks now work in two phases. First, they should first
tell Nix whether they are willing to accept a job. Nix guarantuees
that no two hooks will ever be in the first phase at the same time
(this simplifies the implementation of hooks, since they don't have
to perform locking (?)). Second, if they accept a job, they are
then responsible for building it (on the remote system), and copying
the result back. These can be run in parallel with other hooks and
locally executed jobs.
The implementation is a bit messy right now, though.
* The directory `distributed' shows a (hacky) example of a hook that
distributes build jobs over a set of machines listed in a
configuration file.
parallel as possible (similar to GNU Make's `-j' switch). This is
useful on SMP systems, but it is especially useful for doing builds
on multiple machines. The idea is that a large derivation is
initiated on one master machine, which then distributes
sub-derivations to any number of slave machines. This should not
happen synchronously or in lock-step, so the master must be capable
of dealing with multiple parallel build jobs. We now have the
infrastructure to support this.
TODO: substitutes are currently broken.
This is because the contents of these symlinks are not incorporated
into the hashes of derivations, and could therefore cause a mismatch
between the build system and the target system. E.g., if
`/nix/store' is a symlink to `/data/nix/store', then a builder could
expand this path and store the result. If on the target system
`/nix/store' is not a symlink, or is a symlink that points somewhere
else, we have a dangling pointer.
The trigger for this change is that gcc 3.3.3 does exactly that (it
applies realpath() to some files, such as libraries, which causes
our impurity checker to bail out.)
An annoying side-effect of this change is that it makes it harder to
move the Nix store to a different file system. On Linux, bind
mounts can be used instead of symlink for this purpose (e.g., `mount
-o bind /data/nix/store /nix/store').
writes to stderr:
- `pretty': the old nested style (default)
- `escapes': uses escape codes to indicate nesting and message
level; can be processed using `log2xml'
- `flat': just plain text, no nesting
These can be set using `--log-type TYPE' or the NIX_LOG_TYPE
environment variable.
* Do not create stuff in localstatedir when doing `make install'
(since we may not have write access). In general, installation of
constant code/data should be separate from the initialisation of
mutable state.
chroot() environment.
* A operation `--validpath' to register path validity. Useful for
bootstrapping in a pure Nix environment.
* Safety checks: ensure that files involved in store operations are in
the store.
Nix. This is to prevent Berkeley DB from becoming wedged.
Unfortunately it is not possible to throw C++ exceptions from a
signal handler. In fact, you can't do much of anything except
change variables of type `volatile sig_atomic_t'. So we set an
interrupt flag in the signal handler and check it at various
strategic locations in the code (by calling checkInterrupt()).
Since this is unlikely to cover all cases (e.g., (semi-)infinite
loops), sometimes SIGTERM may now be required to kill Nix.
turned out to be a huge performance bottleneck (the text to printed
would always be evaluated, even when it was above the verbosity
level). This reduces fix-ng execution time by over 50%.
gprof(1) is very useful. :-)
