| Age | Commit message (Collapse) | Author |
|---|
|
|
|
We were incorrectly passing a pointer to a pointer of an
unsigned long long, when we just wanted to pass a pointer
to the unsigned long long. This is especially bad on 32-bit
systems, where we're then writing a 64-bits into a 32-bit value
within ioctl.
We fix this to pass a pointer to the unsigned long long.
Test: On 32-bit device, no longer see native crash from toybox
Bug: http://b/151311535
Signed-off-by: Elliott Hughes <enh@google.com>
|
|
The shift was a remnant from when BLKGETSIZE (which measures in
blocks) was being used on Linux. The Mac has two separate ioctls
for block count and block size, which we're already multiplying
together. And on Linux we're using BLKGETSIZE64, which returns a
result in bytes, not blocks. So lose the shift.
|
|
The recent re-enablement of the BLKGETSIZE64 code broke the Mac
build. Use the equivalent <sys/disk.h> ioctl() pair instead.
|
|
I came here because the new -Wno-unreachable-code-loop-increment warning
didn't like the for loop on line 86. That loop is indeed not necessary.
Use strend() to do a string suffix match.
Use memmem() to search. It's available on macOS and Android by default,
but it's behind _GNU_SOURCE for glibc, so add that to portability.h.
Output the tags in the same order as the Debian modinfo.
I've left "parmtype" in even though the Debian modinfo doesn't output it
at all.
Also fix the tests so that they work on a device that has modules for
multiple kernels installed (like my laptop) --- make sure that the two
modules we pick come from the same kernel.
|
|
(Update to the 64 bit API while we're there. And yes, I checked in the
kernel, it's 512 byte units.)
|
|
Toybox doesn't modify inherited environ[] (the same way we don't modify
our inherited argv[]), so instead of freeing our allocated environ[] when
we want to clear it we need to allocate a new environ[] in the else path
(at a length compatible with the existing plumbing's add stride), set the
first entry of _that_ to 0, and set toys.envc = 1 to record it's an
alloced environ.
|
|
|
|
|
|
Basic readelf(1) implementation, with output close enough to the binutils
version to be usable with scripts that expect the binutils version. This
started as an implementation of nm(1) until I realized that I almost always
want readelf instead, and that you actually have to do much of the work
needed for readelf just to implement nm. Arguably nm (being part of POSIX)
belongs in toybox while readelf doesn't. An argument could also be made that
neither really belongs in toybox, belonging in a separate set of development
tools (such as binutils or the LLVM binutils).
Doesn't support most of the architecture-specific stuff, most notably
relocations, but is aware of things like ARM exidx sections and the common
register state notes in core dumps for the "big four" architectures: arm,
arm64, x86, and x86-64.
Doesn't support symbol versions (but probably should).
Doesn't support section groups or the -t "section details" (which is a long
form of -S "section headers" that I've never seen used in practice and which
isn't part of -a). Doesn't support dumping unwind info or the hash table
bucket histograms.
Reuses the table of ELF architectures from file(1).
Not fuzzed, but successfully parses all the ELF files in my Ubuntu 18.04
system's lib directories. Attempts to exit with an error when presented with
an invalid ELF file rather than struggle on as binutils seems to.
|
|
|
|
|
|
|
|
builtin, and add -u.
|
|
|
|
|
|
An Android engineer complained that they were seeing this when not
running as root:
$ adb shell ls
ls: ./postinstall: Invalid argument
ls: ./init: Permission denied
ls: ./data_mirror: Invalid argument
ls: ./init.environ.rc: Invalid argument
ls: ./metadata: Invalid argument
acct
adb_keys
apex
From strace, it was here:
newfstatat(4, "adb_keys", 0x7fc67eca88, AT_SYMLINK_NOFOLLOW) = -1 EACCES (Permission denied)
readlinkat(4, "adb_keys", 0x5e843c7720, 4095) = -1 EINVAL (Invalid argument)
So stop looking at st.st_mode (and then deciding to do a readlinkat())
if we didn't actually successfully stat().
|
|
|
|
|
|
(Also show unknown values on Linux in hex rather than just "unknown".)
|
|
Major/minor device encoding is not portable. No two BSDs agree with
each other, and Darwin is different again.
Everyone does agree on having major()/minor()/makedev() macros, but
they disagree whether they should be in <sys/types.h> (the BSDs
including Darwin, and old versions of glibc) or <sys/sysmacros.h>
(glibc >= 2.26 and bionic).
This fixes `ls -l /dev/zero` and `stat /dev/zero` on Mac.
|
|
(If DIRTREE_SYMFOLLOW returns ENOENT, skipping the second fstatat() would
also skip the else goto error)
|
|
Includes new tests.
|
|
and #define/#undef a second symbol for the else case.
|
|
|
|
|
|
but --restrict checking should run on the path up to the last component
before unlinking so tar can't be tricked into deleting random files off
the system.
|
|
And a few small cleanups while I was there.
|
|
And check in more multicast support that's been sitting in the tree, I don't
have a test enviornment for it anymore but somebody wanted this...
|
|
(it returns a malloc), and doesn't match the object lifetime of getbasename()
(which always returns some or all of its argument string). The dirname() in
libc modifies its argument string, but that's what posix says to do:
https://pubs.opengroup.org/onlinepubs/9699919799.2008edition/functions/dirname.html
so I guess we can live with it.
|
|
In order to be used as drop in replacement for dirname()
If path is a null pointer or points to an empty string,
dirname() shall return a pointer to the string "." .
|
|
|
|
|
|
|
|
out clang's __has_include(), and && shorts out the eval for gcc.
|
|
The downside is we don't use assembly optimized libc comparison functions,
but the common case is short/no matches until full match. Probably net win.
|
|
|
|
xwaitpid(), fix off by one in xwaitpid().
|
|
The Linux kernel uses the --max-args synonym for -n.
Barbarians who use vi need xargs' -o to be able to do something like:
find -name xargs.c | xargs vi # Sad vi.
find -name xargs.c | xargs -o vi # Happy vi.
The -p option needs fixing to read from /dev/tty because stdin is
otherwise occupied in xargs. I think xargs is the only place that needs
this, so it didn't seem sensible to make all callers to yesno() be
specific about which they wanted, hence the new function.
Also remove the documentation for the build-time XARGS_PEDANTIC option
which isn't actually implemented.
Also add a TODO for -P (which is used by at least one script in the
Linux kernel).
|
|
Tweak DIRTREE_STATLESS so it returns zero stat for any error (I'm testing
that dev, ino, and blksize are all zero), and fill in file type from readdir()
|
|
|
|
GCC 8.2 (which my laptop appears to have been quietly upgraded to)
points out that the %04X of an int might actually take 8 characters;
between that, the "U+", and the trailing NUL, we're gonna need a
bigger boat...
|
|
This was the last user of get_rawline, which lets us remove it.
|
|
ENOENT is ignored, but other errors are reported.
|
|
A TOYFLAG_NOFORK command must run in the context of toysh, but a MAYFORK
can either run standalone or run in the toysh process. MAYFORK means it
cleans up after itself: no leaked resources (malloc, mmap, filehandles, etc),
even in error_exit() paths that would longjmp() back to the shell. It also
doesn't discard anything we need to retain (don't close stdout, change
toys.optargs[] so we can't free it, etc)...
|
|
No current caller except grep needs this, but consistency seems like a
good idea.
Also change the xregcomp error message to be a bit more human-readable,
rather than mention an implementation detail.
|
|
|
|
Test: no EBADF in `strace -e close ./toybox tac /proc/version`
|
|
I started this last night, but thought I'd aim to send multiple small
patches rather than work through all the callers and send one big patch.
I've deliberately chosen the ugly name `allocated_length` because we've
had historical bugs where folks think this a line length in the sense of
the return value. I do wonder whether we should actually have some kind
of getline() wrapper that hides the `char *`/`size_t` pair in lib/,
which makes the function easier to use in most cases but does add the
less common gotcha that you wouldn't be able to getline() through
multiple files at once (which does happen in at least one toy).
But maybe the real fix is to look harder for places where we can just
use loopfiles_lines? Speaking of which, should we actually add two more
arguments to that? Specifically: switch it to getdelim() rather than
getline() behind the scenes, and also add a way to have the trailing
'\n' automatically removed, since that seems to be what most callers
want?
Anyway, that seemed like enough questions that it was time to send this
initial patch out before doing too much more...
|
|
Pointed out during code review of the recent refactor.
|
