kernel/linux.git/drivers/misc/kgdbts.c, branch v4.3.6

clone: support passing tls argument via C rather than pt_regs magic

2015-06-26T00:00:38+00:00

clone has some of the quirkiest syscall handling in the kernel, with a pile of special cases, historical curiosities, and architecture-specific calling conventions. In particular, clone with CLONE_SETTLS accepts a parameter "tls" that the C entry point completely ignores and some assembly entry points overwrite; instead, the low-level arch-specific code pulls the tls parameter out of the arch-specific register captured as part of pt_regs on entry to the kernel. That's a massive hack, and it makes the arch-specific code only work when called via the specific existing syscall entry points; because of this hack, any new clone-like system call would have to accept an identical tls argument in exactly the same arch-specific position, rather than providing a unified system call entry point across architectures. The first patch allows architectures to handle the tls argument via normal C parameter passing, if they opt in by selecting HAVE_COPY_THREAD_TLS. The second patch makes 32-bit and 64-bit x86 opt into this. These two patches came out of the clone4 series, which isn't ready for this merge window, but these first two cleanup patches were entirely uncontroversial and have acks. I'd like to go ahead and submit these two so that other architectures can begin building on top of this and opting into HAVE_COPY_THREAD_TLS. However, I'm also happy to wait and send these through the next merge window (along with v3 of clone4) if anyone would prefer that. This patch (of 2): clone with CLONE_SETTLS accepts an argument to set the thread-local storage area for the new thread. sys_clone declares an int argument tls_val in the appropriate point in the argument list (based on the various CLONE_BACKWARDS variants), but doesn't actually use or pass along that argument. Instead, sys_clone calls do_fork, which calls copy_process, which calls the arch-specific copy_thread, and copy_thread pulls the corresponding syscall argument out of the pt_regs captured at kernel entry (knowing what argument of clone that architecture passes tls in). Apart from being awful and inscrutable, that also only works because only one code path into copy_thread can pass the CLONE_SETTLS flag, and that code path comes from sys_clone with its architecture-specific argument-passing order. This prevents introducing a new version of the clone system call without propagating the same architecture-specific position of the tls argument. However, there's no reason to pull the argument out of pt_regs when sys_clone could just pass it down via C function call arguments. Introduce a new CONFIG_HAVE_COPY_THREAD_TLS for architectures to opt into, and a new copy_thread_tls that accepts the tls parameter as an additional unsigned long (syscall-argument-sized) argument. Change sys_clone's tls argument to an unsigned long (which does not change the ABI), and pass that down to copy_thread_tls. Architectures that don't opt into copy_thread_tls will continue to ignore the C argument to sys_clone in favor of the pt_regs captured at kernel entry, and thus will be unable to introduce new versions of the clone syscall. Patch co-authored by Josh Triplett and Thiago Macieira. Signed-off-by: Josh Triplett Acked-by: Andy Lutomirski Cc: Ingo Molnar Cc: "H. Peter Anvin" Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Thiago Macieira Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

kgdb/kgdbts: support ppc64

2013-03-02T14:52:17+00:00

We can't look up the address of the entry point of the function simply via that function symbol for all architectures. For PPC64 ABI, actually there is a function descriptors structure. A function descriptor is a three doubleword data structure that contains the following values: * The first doubleword contains the address of the entry point of the function. * The second doubleword contains the TOC base address for the function. * The third doubleword contains the environment pointer for languages such as Pascal and PL/1. So we should call a wapperred dereference_function_descriptor() to get the address of the entry point of the function. Note this is also safe for other architecture after refer to "include/asm-generic/sections.h" since: dereference_function_descriptor(p) always is (p) if without arched definition. Signed-off-by: Tiejun Chen Signed-off-by: Jason Wessel

x86,kgdb: Fix DEBUG_RODATA limitation using text_poke()

2012-03-29T22:41:25+00:00

There has long been a limitation using software breakpoints with a kernel compiled with CONFIG_DEBUG_RODATA going back to 2.6.26. For this particular patch, it will apply cleanly and has been tested all the way back to 2.6.36. The kprobes code uses the text_poke() function which accommodates writing a breakpoint into a read-only page. The x86 kgdb code can solve the problem similarly by overriding the default breakpoint set/remove routines and using text_poke() directly. The x86 kgdb code will first attempt to use the traditional probe_kernel_write(), and next try using a the text_poke() function. The break point install method is tracked such that the correct break point removal routine will get called later on. Cc: x86@kernel.org Cc: Thomas Gleixner Cc: Ingo Molnar Cc: H. Peter Anvin Cc: stable@vger.kernel.org # >= 2.6.36 Inspried-by: Masami Hiramatsu Signed-off-by: Jason Wessel

kgdbts: (2 of 2) fix single step awareness to work correctly with SMP

2012-03-29T22:41:24+00:00

The do_fork and sys_open tests have never worked properly on anything other than a UP configuration with the kgdb test suite. This is because the test suite did not fully implement the behavior of a real debugger. A real debugger tracks the state of what thread it asked to single step and can correctly continue other threads of execution or conditionally stop while waiting for the original thread single step request to return. Below is a simple method to cause a fatal kernel oops with the kgdb test suite on a 2 processor ARM system: while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& echo V1I1F100 > /sys/module/kgdbts/parameters/kgdbts Very soon after starting the test the kernel will start warning with messages like: kgdbts: BP mismatch c002487c expected c0024878 ------------[ cut here ]------------ WARNING: at drivers/misc/kgdbts.c:317 check_and_rewind_pc+0x9c/0xc4() [] (check_and_rewind_pc+0x9c/0xc4) [] (validate_simple_test+0x3c/0xc4) [] (run_simple_test+0x1e8/0x274) The kernel will eventually recovers, but the test suite has completely failed to test anything useful. This patch implements behavior similar to a real debugger that does not rely on hardware single stepping by using only software planted breakpoints. In order to mimic a real debugger, the kgdb test suite now tracks the most recent thread that was continued (cont_thread_id), with the intent to single step just this thread. When the response to the single step request stops in a different thread that hit the original break point that thread will now get continued, while the debugger waits for the thread with the single step pending. Here is a high level description of the sequence of events. cont_instead_of_sstep = 0; 1) set breakpoint at do_fork 2) continue 3) Save the thread id where we stop to cont_thread_id 4) Remove breakpoint at do_fork 5) Reset the PC if needed depending on kernel exception type 6) soft single step 7) Check where we stopped if current thread != cont_thread_id { if (here for more than 2 times for the same thead) { ### must be a really busy system, start test again ### goto step 1 } goto step 5 } else { cont_instead_of_sstep = 0; } 8) clean up and run test again if needed 9) Clear out any threads that were waiting on a break point at the point in time the test is ended with get_cont_catch(). This happens sometimes because breakpoints are used in place of single stepping and some threads could have been in the debugger exception handling queue because breakpoints were hit concurrently on different CPUs. This also means we wait at least one second before unplumbing the debugger connection at the very end, so as respond to any debug threads waiting to be serviced. Cc: stable@vger.kernel.org # >= 3.0 Signed-off-by: Jason Wessel

kgdbts: (1 of 2) fix single step awareness to work correctly with SMP

2012-03-29T22:41:24+00:00

The do_fork and sys_open tests have never worked properly on anything other than a UP configuration with the kgdb test suite. This is because the test suite did not fully implement the behavior of a real debugger. A real debugger tracks the state of what thread it asked to single step and can correctly continue other threads of execution or conditionally stop while waiting for the original thread single step request to return. Below is a simple method to cause a fatal kernel oops with the kgdb test suite on a 4 processor x86 system: while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& while [ 1 ] ; do ls > /dev/null 2> /dev/null; done& echo V1I1F1000 > /sys/module/kgdbts/parameters/kgdbts Very soon after starting the test the kernel will oops with a message like: kgdbts: BP mismatch 3b7da66480 expected ffffffff8106a590 WARNING: at drivers/misc/kgdbts.c:303 check_and_rewind_pc+0xe0/0x100() Call Trace: [] check_and_rewind_pc+0xe0/0x100 [] validate_simple_test+0x25/0xc0 [] run_simple_test+0x107/0x2c0 [] kgdbts_put_char+0x18/0x20 The warn will turn to a hard kernel crash shortly after that because the pc will not get properly rewound to the right value after hitting a breakpoint leading to a hard lockup. This change is broken up into 2 pieces because archs that have hw single stepping (2.6.26 and up) need different changes than archs that do not have hw single stepping (3.0 and up). This change implements the correct behavior for an arch that supports hw single stepping. A minor defect was fixed where sys_open should be do_sys_open for the sys_open break point test. This solves the problem of running a 64 bit with a 32 bit user space. The sys_open() never gets called when using the 32 bit file system for the kgdb testsuite because the 32 bit binaries invoke the compat_sys_open() call leading to the test never completing. In order to mimic a real debugger, the kgdb test suite now tracks the most recent thread that was continued (cont_thread_id), with the intent to single step just this thread. When the response to the single step request stops in a different thread that hit the original break point that thread will now get continued, while the debugger waits for the thread with the single step pending. Here is a high level description of the sequence of events. cont_instead_of_sstep = 0; 1) set breakpoint at do_fork 2) continue 3) Save the thread id where we stop to cont_thread_id 4) Remove breakpoint at do_fork 5) Reset the PC if needed depending on kernel exception type 6) if (cont_instead_of_sstep) { continue } else { single step } 7) Check where we stopped if current thread != cont_thread_id { cont_instead_of_sstep = 1; goto step 5 } else { cont_instead_of_sstep = 0; } 8) clean up and run test again if needed Cc: stable@vger.kernel.org # >= 2.6.26 Signed-off-by: Jason Wessel

kgdbts: Fix kernel oops with CONFIG_DEBUG_RODATA

2012-03-29T22:41:24+00:00

On x86 the kgdb test suite will oops when the kernel is compiled with CONFIG_DEBUG_RODATA and you run the tests after boot time. This is regression has existed since 2.6.26 by commit: b33cb815 (kgdbts: Use HW breakpoints with CONFIG_DEBUG_RODATA). The test suite can use hw breakpoints for all the tests, but it has to execute the hardware breakpoint specific tests first in order to determine that the hw breakpoints actually work. Specifically the very first test causes an oops: # echo V1I1 > /sys/module/kgdbts/parameters/kgdbts kgdb: Registered I/O driver kgdbts. kgdbts:RUN plant and detach test Entering kdb (current=0xffff880017aa9320, pid 1078) on processor 0 due to Keyboard Entry [0]kdb> kgdbts: ERROR PUT: end of test buffer on 'plant_and_detach_test' line 1 expected OK got $E14#aa WARNING: at drivers/misc/kgdbts.c:730 run_simple_test+0x151/0x2c0() [...oops clipped...] This commit re-orders the running of the tests and puts the RODATA check into its own function so as to correctly avoid the kernel oops by detecting and using the hw breakpoints. Cc: # >= 2.6.26 Signed-off-by: Jason Wessel

drivers/misc: Add module.h to files who are really modular.

2011-10-31T23:31:53+00:00

These files really need the full module.h header file present, but were just getting it implicitly before. Fix it up in advance so we avoid build failures once the cleanup commit is present. Signed-off-by: Paul Gortmaker

kgdbts: only use new asm-generic/ptrace.h api when needed

2011-06-01T07:35:01+00:00

The new instruction_pointer_set helper is defined for people who have converted to asm-generic/ptrace.h, so don't use it generally unless the arch needs it (in which case it has been converted). This should fix building of kgdb tests for arches not yet converted. Signed-off-by: Mike Frysinger Acked-by: Stephen Rothwell Cc: Jason Wessel Cc: Andrew Morton Signed-off-by: Linus Torvalds

kgdbts: unify/generalize gdb breakpoint adjustment

2011-05-27T00:12:36+00:00

The Blackfin arch, like the x86 arch, needs to adjust the PC manually after a breakpoint is hit as normally this is handled by the remote gdb. However, rather than starting another arch ifdef mess, create a common GDB_ADJUSTS_BREAK_OFFSET define for any arch to opt-in via their kgdb.h. Signed-off-by: Mike Frysinger Cc: Oleg Nesterov Cc: Jason Wessel Cc: Thomas Gleixner Cc: Ingo Molnar Cc: "H. Peter Anvin" Acked-by: Paul Mundt Acked-by: Dongdong Deng Cc: Sergei Shtylyov Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

Fix common misspellings

2011-03-31T14:26:23+00:00

Fixes generated by 'codespell' and manually reviewed. Signed-off-by: Lucas De Marchi