Fix OOPS caused by reference of config pointer

The config pointer points to data that may vanish across the life-time
of the ring buffer stream references. It's only kept valid for the
duration between the channel create and destroy, but the streams can
keep an extra reference on the channel, and thus need the config to stay
valid.

Perform a copy of the config rather than keeping a pointer to it.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Gather detailed info from x86 64 32-bit compat syscall instrumentation

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng lib: ring buffer move null pointer check to open

Let's move the NULL buf check to the file "open", where it belongs.  The
"open" file operation is the actual interface between lib ring buffer
and the modules using it.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng lib: ring buffer remove duplicate null pointer

This check is performed at mapping setup time in
lib_ring_buffer_mmap_buf() already, so we can safely remove this
duplicata.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng lib: ring buffer: remove stale null-pointer

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng wrapper: add missing include to kallsyms wrapper

Needed to keep bissectability.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng: cleanup one-bit signed bitfields

* Dan Carpenter <dan.carpenter@oracle.com> wrote:
> Sparse complains that these signed bitfields look "dubious".  The
> problem is that instead of being either 0 or 1 like people would expect,
> signed one bit variables like this are either 0 or -1.  It doesn't cause
> a problem in this case but it's ugly so lets fix them.

* walter harms (wharms@bfs.de) wrote:
> hi,
> This patch looks ok to me but this design is ugly by itself.
> It should be replaced by an uchar uint whatever or use a
> real bool (obviously not preferred by this programmes).

bool :1, uchar :1 or uint :1 could make sense. uchar:1/bool:1 won't save
any space here, because the surrounding fields are either uint or
pointers, so alignment will just add padding.

I try to use int/uint whenever possible because x86 CPUs tend to get
less register false-dependencies when using instructions modifying the
whole register (generated by using int/uint types) rather than only part
of it (uchar/char/bool). I only use char/uchar/bool when there is a
clear wanted space gain.

The reason why I never use the bool type within a structure when I want
a compact representation is that bool takes a whole byte just to
represent one bit:

struct usebitfield {
int a;
unsigned int f:1, g:1, h:1, i:1, j:1;
int b;
};

struct usebool {
int a;
bool f, g, h, i, j;
int b;
};

struct useboolbf {
int a;
bool f:1, g:1, h:1, i:1, j:1;
int b;
};

int main()
{
printf("bitfield %d bytes, bool %d bytes, boolbitfield %d bytes\n",
sizeof(struct usebitfield), sizeof(struct usebool),
sizeof(struct useboolbf));
}

result:

bitfield 12 bytes, bool 16 bytes, boolbitfield 12 bytes

This is because each bool takes one byte, while the bitfields are put in
units of "unsigned int" (or bool for the 3rd struct). So in this
example, we need 5 bytes + 3 bytes alignment for the bool, but only 4
bytes to hold the "unsigned int" unit for the bitfields.

The choice between bool and bitfields must also take into account the
frequency of access to the variable, because bitfields require mask
operations to access the selected bit(s). You will notice that none of
these bitfields are accessed on the tracing fast-path: only in
slow-paths. Therefore, space gain is more important than speed here.

One might argue that I have so few of these fields here that it does not
make an actual difference to go for bitfield or bool. I am just trying
to choose types best suited for their intended purpose, ensuring they
are future-proof and will allow simply adding more fields using the same
type, as needed.

So I guess I'll go for unsigned int :1 (rather than uint:1, so we
keep source-level compatibility with user-space LTTng-UST).

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add TODO file

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update symbol name length max size to 256

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix last modifications to string_from_user operations

Signed-off-by: Yannick Brosseau <yannick.brosseau@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Document that depmod needs to be executed by hand

External modules don't seem to actually execute depmod on make install,
even though the console output says so. Document that it needs to be
manually executed in the README file.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix strlen_user fault space reservation

The previous fix took care of kernel OOPS, but did not reserve space for
the \0.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix tp_copy_string_from_user handling of faults

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Disable block layer tracing support for kernels < 2.6.38

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng context: perf counter, fix 32-bit vs 64-bit field size bug

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Update trace clock warning to match the current development plan

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ringbuffer: make ring buffer printk less verbose

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Makefile: do not run depmod manually

The modules_install target of the kernel build infrastructure already
runs depmod properly, taking into account the cross-compilation
case. Therefore, it is useless to re-run depmod here, and also harmful
since it does not work in cross-compilation contexts (it does the
depmod on the modules installed on the host and not the ones being
built for the target).

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 2.0-pre10

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

x86-32: override sys_mmap syscall instrumentation

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix cast warning

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Version 2.0-pre9

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

x86 32 syscalls: fix !CONFIG_UID16

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use __string_from_user

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add __string_from_user

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Complete comm -> procname change

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Rename "comm" context to "procname"

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fetch "type" name field for mount syscall

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Create common override file for all architectures

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Shrink stack size usage

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Cleanup: remove unused lttng-compat-syscalls.c file

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Support detailed compat syscalls on x86_64

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix syscall table

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add disabled compat syscalls detail (work in progress)

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add config compat support for detailed x86_32 syscalls

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Handle newer 3.1+ perf API

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add dummy x86-32 override headers

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Bind x86-32 syscall headers

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add x86-32 syscall headers

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add check for KALLSYMS_ALL in syscall extraction script

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Automate string extraction for syscalls

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Automate syscall filename/pathname extraction

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add exit to ring_buffer_frontend so module is unloadable

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add sys_open filename (x86_64)

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add missing memcpy callback

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

copy_from_user and memset

This patch provides the copy_from_user and memset operations for the lib
ringbuffer.

[ edit: tp_copy_string_from_user len parameter now excludes final \0 ]

Signed-off-by: Julien Desfossez <julien.desfossez@polymtl.ca>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add missing alignment after header write

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lib ring buffer: add frontend init as module_init

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lib ring buffer frontend: add missing lock initialization

[...]
while enabling the events, I get following output:

[ 4304.827819] BUG: spinlock bad magic on CPU#1, ltt-sessiond/1305
[ 4304.841735]  lock: 81453c00, .magic: 00000000, .owner: <none>/-1,
.owner_cpu: 0
[ 4304.857421] [<800110ad>] (unwind_backtrace+0x1/0x90) from
[<8022c847>] (do_raw_spin_lock+0x17/0xfc)
[ 4304.875213] [<8022c847>] (do_raw_spin_lock+0x17/0xfc) from
[<7f97c30f>] (channel_create+0xd6/0x158 [lib_ring_buffer])
[ 4304.894866] [<7f97c30f>] (channel_create+0xd6/0x158
[lib_ring_buffer]) from [<7f99856b>] (_channel_create+0x28/0x32
[ltt_ring_buffer_client_discard])
[ 4304.917755] [<7f99856b>] (_channel_create+0x28/0x32
[ltt_ring_buffer_client_discard]) from [<7f98b507>]
(ltt_channel_create+0x12a/0x190 [ltt_relay])
[ 4304.940551] [<7f98b507>] (ltt_channel_create+0x12a/0x190 [ltt_relay])
from [<7f98c34b>] (lttng_abi_create_channel+0xba/0x158 [ltt_relay])
[ 4304.962341] [<7f98c34b>] (lttng_abi_create_channel+0xba/0x158
[ltt_relay]) from [<7f98c43f>] (lttng_session_ioctl+0x56/0x64 [ltt_relay])
[ 4304.984436] [<7f98c43f>] (lttng_session_ioctl+0x56/0x64 [ltt_relay])
from [<800baa49>] (do_vfs_ioctl+0x1a5/0x1dc)
[ 4305.004394] [<800baa49>] (do_vfs_ioctl+0x1a5/0x1dc) from [<800baab3>]
(sys_ioctl+0x33/0x4c)
[ 4305.022369] [<800baab3>] (sys_ioctl+0x33/0x4c) from [<8000c2c1>]
(ret_fast_syscall+0x1/0x50)
All kernel events are enabled in channel channel0

Reported-by: Avik Sil <avik.sil@linaro.org>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ARM: Set bit 0 for thumb mode in kallsyms_lookup_name returned address (missing file)

Adding missing file in commit c539a324b020171c10d0fcb100892ac93c17e9e7

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ARM: Set bit 0 for thumb mode in kallsyms_lookup_name returned address

This patch fixes the undefined instruction oops due to execution
of thumb-2 code in ARM mode. The zero bit in the symbol address
returned by kallsyms_lookup_name is not set, leading to switching
to ARM mode that generates oops while executing thumb-2 code. For
detailed discussion, see [1].
[1] http://lists.casi.polymtl.ca/pipermail/ltt-dev/2011-September/005176.html

v1:
        - include wrapper function kallsyms_lookup_funcptr as suggested
        by Dave Martin

Signed-off-by: Avik Sil <avik.sil@linaro.org>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Create override for sys_mmap

based on mmap(2).

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Rename syscalls -> syscall in LTTng ABI

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use different event name for compat syscalls

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Allow trace syscall table override

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add dummy override files

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

syscall detail: add override capability

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add pointer pretty-print heuristics

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add __field_hex

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Rename sys exit errno into ret

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Enable detailed syscalls only for native x86_64

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use "unknown" for compat syscalls

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Disable x86-64 specific detailed syscalls for now, until we deal with 32-bit compat

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix syscall exit

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix syscall exit tracing

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Remove unused variable

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix sys exit

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix unknown syscall support

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Only enable detailed syscalls for x86_64 for now

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Merge unknown syscall method with extended detail method

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add underscore prefix before event names

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add syscalls pointers detailed syscall tracing

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Use perl for regexp

Allows more than 9 backward refs.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Support generation of syscall probes for those with pointers

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Support detailed syscalls with 0 arguments

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

detailed syscall tracing (work in progress)

Buidl disabled for now.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Expose /proc/lttng file

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Implement detailed syscall event probe

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Disable perf build for < 2.6.33 kernels

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Syscall detail mode: generate syscall table

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

block instrumentation: write cmd as hex sequence

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng-events TRACE_EVENT generator cleanup

Adds support for _hex variants while cleaning up and refactoring macros,
removing duplicatas.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

block layer instrumentation: fix cmd field

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Remove 0 args syscalls for now from type-aware syscall cb generator

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Generate trace syscall table macros

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

standardize on "syscalls"

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Put syscall tracing without arg types back in place

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Revert to working code for now

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Detailed system call tracing (TRACE_EVENT() generation for integer support)

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

lttng events probes: don't clash with "ctx" variable name

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Merge branch 'master' of ssh://git.lttng.org/home/git/lttng-modules

lib ring buffer must take private pointer copy before decrementing refcount

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

ChangeLog for v2.0-pre8

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix metadata buffer wait/wakeup

The metadata producer should be waiting on its own wait queue, woken up
when the consumer position is moved by the consumer.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add changelog

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add missing module refcount to lttng_tracepoint_list_fops

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Add license info about lib/ringbuffer/ring_buffer_mmap.c being GPLv2

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Fix calibrate: add noinline

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Calibrate namespacing

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Merge branch 'master' of ssh://git.lttng.org/home/git/lttng-modules