| Age | Commit message (Collapse) | Author |
|---|
|
|
|
In issue #5554 Michael wrote:
The implementation of sigaction on i386 for older kernels makes the system call using an inline asm element with two flaws:
1. The asm is not marked as depending on the kact structure or modifying the koact structure. Thus, GCC is free to assume these structures need not be kept consistent, allowing it to remove all initialization of kact.
2. The asm allows the signal number to be provided as a memory reference. But this allows GCC to provide a stack-relative operand, which will break because the assembler saves %ebx on the stack before using that operand.
1 didn't use to be a problem in practice because GCC 4.2.* didn't seize the optimization opportunity. GCC 4.3.2, however, optimizes out the "kact.sa_flags = act->sa_flags | SA_RESTORER;" line, so that the kernel sees garbage in sa_flags. This can result in the kernel seeing the SA_RESETHAND flag, causing erratic behaviour in signal dependent programs.
2 becomes an issue if "-fomit-frame-pointer" is provided. In uClibc-0.9.29 it isn't, uClibc-0.9.30-rc2 does use the flag by default.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Tested successfully on i386..
|
|
|
|
|
|
|
|
like
o UCLIBC_HAS_GNU_ERROR
o UCLIBC_HAS_BSD_ERR
o UCLIBC_HAS_PTY
o UCLIBC_HAS_GETPT (1)
o UCLIBC_SYSCALL_STUBS
o UCLIBC_SYSCALL_STUB_WARNING
o UCLIBC_LINUX_SPECIFIC (2)
o UCLIBC_BSD_SPECIFIC (3)
o UCLIBC_NTP_LEGACY (4)
o UCLIBC_SV4_DEPRECATED (5)
o UCLIBC_HAVE_REALTIME (6)
o UCLIBC_HAVE_ADVANCED_REALTIME (7)
o UCLIBC_HAVE_EPOLL (8)
o UCLIBC_HAVE_XATTR (9)
o UCLIBC_HAVE_PROFILING (10)
(1) make non-standard getpt optional and implement standard posix_openpt
(2) fstatfs(), inotify_*(), ioperm(), iopl(), madvise(), modify_ldt(),
personality()
ppoll(), setresuid()
(3) mincore(), getdomainname(), setdomainname()
(4) ntp_adjtime(), ntp_gettime() aliases
(5) ustat() [use statfs(2) in your code instead]
(6) All marked as "(REALTIME)" in SUSv3
(7) All marked as "(ADVANCED REALTIME)" in SUSv3
(8) epoll_create(), epoll_ctl(), epoll_wait()
(9) all Extended Attributes
(10) helpers for gcc's -finstrument-functions
- Fixes _dl_exit()
- Implements sleep(3) for !UCLIBC_HAVE_REALTIME
- Implements usleep(3) for !UCLIBC_HAVE_REALTIME
- adds #warning about incorrect posix_fadvise{,64}()
- removes unused and unwanted uselib()
Net outcome is that an allnoconfig with HAVE_SHARED is now about 88k instead
of formerly 130k.
|
|
even if arch overrides smallint size. Makes it easier to spot bugs
on such arches.
|
|
- 79 0 28 107 6b libc/inet/rpc/create_xid.o
+ 76 0 25 101 65 libc/inet/rpc/create_xid.o
- 126 0 4 130 82 libc/misc/assert/__assert.o
+ 123 0 1 124 7c libc/misc/assert/__assert.o
- 648 4 24 676 2a4 libc/misc/internals/__uClibc_main.o
+ 645 4 21 670 29e libc/misc/internals/__uClibc_main.o
- 230 0 4 234 ea libc/stdlib/abort.o
+ 216 0 1 217 d9 libc/stdlib/abort.o
- 129 0 4 133 85 libc/termios/tcgetsid.o
+ 126 0 1 127 7f libc/termios/tcgetsid.o
|
|
in string.h and strings.h. This caught unguarded string ops in
libc/inet/ethers.c __ether_line_w() function.
I will wait for fallout reports for a week or so,
then continue converting more libc_hidden_proto's.
|
|
|
|
- synch F_LINUX_SPECIFIC_BASE related fcntls for all arches
|
|
- pull bits/fcntl.h from glibc (Timo Teräs)
|
|
|
|
a problem where the linker was trying to use the wrong symbol name for the
init function.
Define SYMBOL_PREFIX as _ in Rules.mak for h8300, bfin, i960,
microblaze, and v850. Add -D__UCLIBC_UNDERSCORES__ in CFLAGS for targets
which define SYMBOL_PREFIX as _. Remove defines and undefs from
uClibc_arch_features.h of each target.
Add $(SYMBOL_PREFIX) to __uClibc_init when passed by ld option -init.
|
|
doing double negatives
|
|
On an i386 platform with no rt_sigsuspend syscall (ie: Linux 2.0), compilation will halt on libc/sysdeps/linux/common/sigsuspend.os with a cryptic error message:
"Error: non-constant expression in ".if" statement"
I've investigated and found that the cause is that a literal '0' is being passed into a block of complex assembler macrology that is only prepared to deal with register names - '%eax', etc.
In turn, that seems to be because of a typo in the GCC register constraints. The constraints for 2 and 3-argument syscalls includes a "C" constraint. To gcc, "C" means an SSE floating point constant -- an unlikely element in a syscall. I suspect the author meant to type "S" (%esi).
|
|
from i386 and x86_64.
|
|
else.
- internal_function is only used inside libc, so do not define it in
uClibc_arch_features.h on i386 if we are outside of libc.
|
|
with gcc 4.2. Renaming to make it a bit uncommon name.
|
|
Import INTERNAL_SYSCALL macro for i386 from glibc.
Use above macro in posix_fadvise implementation if an arch defines it.
|
|
|
|
applications stop using _syscall#() and use syscall() instead. Cleanup
internal handling of syscall includes to use the correct header file.
|
|
minor include file issues
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
based profiling I nuked ages ago since tools like oprofile are non invasive
and work so much better.
|
|
|
|
|
|
|
|
|
|
most of global data relocations are back
|
|
|
|
|
|
the archs fail
|
|
call, and dont bother with 64bit versions on 64bit hosts as the regular one works fine (should fix the setrlimit ltp tests)
|
|
to be weaks
|
|
|
