get_loadavg doesn't work as intended. According to the comments, it
should be returning an average over a few seconds, but it is actually
reading the instantaneous load. It is almost always returning 0, but
can sometimes, depending on workload, spike very high into the hundreds
even when the average cpu load is under 10%. Disable it for now.
Change-Id: I63ed100af1cf9463549939b8113ed83676db5f86
Signed-off-by: Colin Cross <ccross@android.com>
Decare war on uninterruptible sleep. Add a tracepoint which
walks the kernel stack and dumps the first non-scheduler function
called before the scheduler is invoked.
Change-Id: I19e965d5206329360a92cbfe2afcc8c30f65c229
Signed-off-by: Riley Andrews <riandrews@google.com>
Dumps registers and stacktrace into console-ramoops when called
from a watchdog fiq.
Change-Id: Ib6fab5a52f670db18e64214d5e4890e8292a749c
Signed-off-by: Arve Hjønnevåg <arve@android.com>
This allows the output from the register and stack trace code to be
sent elsewhere.
Change-Id: I41bb0d5a25e1b9ca55feef5dbd675818b2f832d5
Signed-off-by: Arve Hjønnevåg <arve@android.com>
Add fiq_debugger_arm64.c that implements the platform-specific
functions.
Change-Id: I4d8b96777bb8503a93d4eb47bbde8e018740a5bf
Signed-off-by: Colin Cross <ccross@android.com>
Split arm support into a separate .c file that is only built for
CONFIG_ARM.
Change-Id: Iba16f4d51608bf9c3e5c8acefefcd38fead9797c
Signed-off-by: Colin Cross <ccross@android.com>
IRQ mode already passes in a struct pt_regs from get_irq_regs().
FIQ mode passes in something similar but not identical to a
struct pt_regs - FIQ mode stores the spsr of the interrupted mode
in slot 17, while pt_regs expects orig_r0.
Replace the existing mixture of void *regs, unsigned *regs, and
struct pt_regs * const with const struct pt_regs *. Modify
dump_regs not to print the spsr since it won't be there in a
struct pt_regs anyways. Modify dump_allregs to highlight the
mode that was interrupted, making spsr easy to find there.
Change-Id: Ibfe1723d702306c7605fd071737d7be9ee9d8c12
Signed-off-by: Colin Cross <ccross@android.com>
Allow compiling fiq_debugger.c without CONFIG_FIQ_GLUE for
platforms that don't support FIQs.
Change-Id: Iabdfd790d24fa9d47b29d2f850c567af2dcad78f
Signed-off-by: Colin Cross <ccross@android.com>
Rename variables and functions in the global namespace to avoid
future collisions.
Change-Id: Ic23a304b0f794efc94cc6d086fddd63231d99c98
Signed-off-by: Colin Cross <ccross@android.com>
Move fiq_debugger into drivers/staging/android/fiq_debugger/ to
allow for sharing between ARM and ARM64.
Change-Id: I6ca5e8b7e3d000f57da3234260261c5592cef2a8
Signed-off-by: Colin Cross <ccross@android.com>
Avoid conflicts with user mode usage of the same instructions, as with
Clang -ftrapv.
Change-Id: I12d1c6d8f94376bfd2503cb0be843d7e478fb6ea
Signed-off-by: Todd Poynor <toddpoynor@google.com>
debug_console_write calls debug_uart_flush, which will usually wait
until the serial port fifo empties. If another thread is continuously
calling fiq_tty_write, the fifo will constantly be refilled and
debug_uart_flush might never return.
Add a spinlock that is locked in debug_console_write and fiq_tty_write
to ensure they can't run at the same time. This has an extra advantage
of preventing lines from the console and tty from being mixed together.
Also reduce the size returned by fiq_tty_write_room to keep the time
spent with the spinlock held to a reasonable value.
In addition, make sure fiq context can't loop forever by never calling
debug_uart_flush when the console is enabled.
Change-Id: I5712b01f740ca0c84f680d2032c9fa16b7656939
Signed-off-by: Colin Cross <ccross@android.com>
kernel_restart cannot be called from interrupt context. Add support for
commands called from a work function, and implement the "reboot" command
there. Also rename the existing irq-mode command to "reset" and change
it to use machine_restart instead of kernel_restart.
Change-Id: I3c423147c01db03d89e95a5b99096ca89462079f
Signed-off-by: Colin Cross <ccross@android.com>
Fix setting up consoles on multiple fiq debugger devices by
splitting the tty driver init into the initcall, and initializing
the single tty device during probe. Has the side effect of moving
the tty device node to /dev/ttyFIQx, where x is the platform device
id, which should normally match the serial port.
To avoid having to pass a different console=/dev/ttyFIQx for every
device, make the fiq debugger a preferred console that will be used
by default if no console was passed on the command line.
Change-Id: I6cc2670628a41e84615859bc96adba189966d647
Signed-off-by: Colin Cross <ccross@android.com>
kdb expects carriage returns through the serial port to terminate
commands. Modify it to accept the first seen carriage return or
new line as a terminator, but not treat \r\n as two terminators.
Change-Id: I06166017e7703d24310eefcb71c3a7d427088db7
Signed-off-by: Colin Cross <ccross@android.com>
Adds polling tty ops to the fiq debugger console tty, which allows
kgdb to run against an fiq debugger console.
Add a check in do_sysrq to prevent enabling kgdb from the fiq
debugger unless a flag (writable only by root) has been set. This
should make it safe to enable KGDB on a production device.
Also add a shortcut to enable the console and kgdb together, to
allow kgdb to be enabled when the shell on the console is not
responding.
Change-Id: Ifc65239ca96c9887431a6a36b9b44a539002f544
Signed-off-by: Colin Cross <ccross@android.com>
Convert all the calls to state->pdata->uart_putc to a debug_putc
helper.
Change-Id: Idc007bd170ff1b51d0325e238105ae0c86d23777
Signed-off-by: Colin Cross <ccross@android.com>
Pass the rest of the reboot command to kernel_restart to allow
reboot bootloader to work from FIQ debugger.
Change-Id: I4e7b366a69268dda17ffcf4c84f2373d15cb1271
Signed-off-by: Colin Cross <ccross@android.com>
Call kernel_restart instead of arch_reset, the ARM reset handling
has changed.
Remove localtimer irq printing, they now show up in the regular
irq stats.
Change-Id: I523da343b292c5711f3e1cbfd766d32eea2da84e
Signed-off-by: Colin Cross <ccross@android.com>
- In the current implementation, when a signal is sent to the reading process,
read is cancelled by calling usb_ep_dequeue, which lead into calling
acc_complete_out with ECONNRESET, but the current logic treats it as
disconnection, which makes the device inaccessible until cable is actually
disconnected.
- The fix calls disconnect only when ESHUTDOWN error is passed.
- If data has already arrived while trying cancelling, the data is marked
as available, and it will be read out on the next read. This is necessary
as USB bulk is assumed to guarantee no data loss.
Signed-off-by: keunyoung <keunyoung@google.com>
Use '%zd' format specifier for size_t type instead
of '%d' to fix build warnings like:
drivers/usb/gadget/function/u_ether.c: In function ‘rx_submit’:
drivers/usb/gadget/function/u_ether.c:244:2: warning: format ‘%d’ expects argument of type ‘int’, but argument 4 has type ‘size_t’ [-Wformat=]
DBG(dev, "%s: size: %d\n", __func__, size);
^
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Use rndis_params instead of configNr to align with
changes from mainline commit 83210e59ee
"usb: gadget: rndis: use rndis_params instead of configNr".
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Declare "cdev" to fix broken AOSP commit
"RNDIS: Add Data aggregation (multi packet) support",
otherwise we run into following build failure:
CC drivers/usb/gadget/function/f_rndis.o
drivers/usb/gadget/function/f_rndis.c: In function ‘rndis_command_complete’:
drivers/usb/gadget/function/f_rndis.c:479:3: error: ‘cdev’ undeclared (first use in this function)
drivers/usb/gadget/function/f_rndis.c:479:3: note: each undeclared identifier is reported only once for each function it appears in
make[4]: *** [drivers/usb/gadget/function/f_rndis.o] Error 1
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Use '%z' format specifier for sizeof operator instead
of '%u' to fix build warnings like:
warning: format '%u' expects type 'unsigned int', but argument 3 has type 'long unsigned int'
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Remove unused variables and functions to fix following
build warnings:
CC drivers/usb/gadget/configfs.o
drivers/usb/gadget/configfs.c: In function ‘gadgets_make’:
drivers/usb/gadget/configfs.c:1710:6: warning: unused variable ‘err’ [-Wunused-variable]
int err;
^
drivers/usb/gadget/configfs.c:1709:27: warning: unused variable ‘attr’ [-Wunused-variable]
struct device_attribute *attr;
^
drivers/usb/gadget/configfs.c:1708:28: warning: unused variable ‘attrs’ [-Wunused-variable]
struct device_attribute **attrs;
^
drivers/usb/gadget/configfs.c: In function ‘gadgets_drop’:
drivers/usb/gadget/configfs.c:1774:27: warning: unused variable ‘attr’ [-Wunused-variable]
struct device_attribute *attr;
^
drivers/usb/gadget/configfs.c:1773:28: warning: unused variable ‘attrs’ [-Wunused-variable]
struct device_attribute **attrs;
^
<snip>...
CC drivers/usb/gadget/function/f_mtp.o
drivers/usb/gadget/function/f_mtp.c:1219:12: warning: ‘mtp_bind_config’ defined but not used [-Wunused-function]
static int mtp_bind_config(struct usb_configuration *c, bool ptp_config)
^
drivers/usb/gadget/function/f_mtp.c:1300:12: warning: ‘mtp_setup’ defined but not used [-Wunused-function]
static int mtp_setup(void)
^
<snip>...
CC drivers/usb/gadget/function/f_accessory.o
drivers/usb/gadget/function/f_accessory.c:969:1: warning: ‘acc_function_bind’ defined but not used [-Wunused-function]
acc_function_bind(struct usb_configuration *c, struct usb_function *f) {
^
drivers/usb/gadget/function/f_accessory.c:1172:12: warning: ‘acc_bind_config’ defined but not used [-Wunused-function]
static int acc_bind_config(struct usb_configuration *c)
^
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Also select SND_PCM while building f_audio_source
otherwise we run into following build error:
LD init/built-in.o
drivers/built-in.o: In function `audio_data_complete':
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:458: undefined reference to `snd_pcm_period_elapsed'
drivers/built-in.o: In function `audio_pcm_hw_free':
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:770: undefined reference to `snd_pcm_lib_free_vmalloc_buffer'
drivers/built-in.o: In function `snd_pcm_lib_alloc_vmalloc_buffer':
/linaro/android/kernel/linaro-android/include/sound/pcm.h:1179: undefined reference to `_snd_pcm_lib_alloc_vmalloc_buffer'
drivers/built-in.o: In function `audio_pcm_open':
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:734: undefined reference to `snd_pcm_limit_hw_rates'
drivers/built-in.o: In function `snd_card_setup':
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:888: undefined reference to `snd_pcm_new'
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:898: undefined reference to `snd_pcm_set_ops'
/linaro/android/kernel/linaro-android/drivers/usb/gadget/function/f_audio_source.c:899: undefined reference to `snd_pcm_lib_preallocate_pages_for_all'
drivers/built-in.o:(.data+0x1fd28): undefined reference to `snd_pcm_lib_ioctl'
make: *** [vmlinux] Error 1
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
There is this new mandatory UDC->reset API in v3.18+ kernels,
commit ef979a26 "usb: gadget: add reset API at usb_gadget_driver".
Let android_disconnect handle that for Android, similar to
how composite_disconnect is handling the generic ConfigFS
gadget reset request.
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Android frameworks relies on the alsa
config reported by the f_midi device.
Signed-off-by: Badhri Jagan Sridharan <Badhri@google.com>
Change-Id: I0695e00b166fd953f50acea93802245b0d5a5240
Android frameworks (UsbDeviceManager) relies on gadget state exported
through device attributes. This CL adds the device attribute to export
USB gadget state.
Change-Id: Id0391810d75b58c579610fbec6e37ab22f28886d
Signed-off-by: Badhri Jagan Sridharan <Badhri@google.com>
Add a missing ";" after EXPORT_SYMBOL() otherwise we run
into following build error if Kernel Modules are supported:
----------
CC drivers/usb/phy/class-dual-role.o
drivers/usb/phy/class-dual-role.c:91:1: error: expected ',' or ';' before 'int'
int dual_role_get_property(struct dual_role_phy_instance *dual_role,
^
make[3]: *** [drivers/usb/phy/class-dual-role.o] Error 1
----------
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
This CL adds a new class to monitor and change
dual role usb ports from userspace. The usb
phy drivers can register to the dual_role_usb
class and expose the capabilities of the ports.
The phy drivers can decide on whether a specific
attribute can be changed from userspace by
choosing to implement the appropriate callback.
Cherry-picked from
https://android-review.googlesource.com/#/c/167310/
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
Bug: 21615151
Change-Id: Id1c4aaa97e898264d7006381a7badd029b5d9789
Fix NULL pointer dereference while trying to link PTP
function to a gadget configuration without creating
MTP function.
PTP piggyback on MTP function so make sure we have
MTP function created beforehand. Otherwise we run
into following kernel panic:
-----------------------
[ 70.329957] Unable to handle kernel NULL pointer dereference at virtual address 00000000
[ 70.330738] pgd = dd8ec000
[ 70.330916] [00000000] *pgd=00000000
[ 70.331663] Internal error: Oops: 805 [#1] SMP THUMB2
[ 70.332155] CPU: 0 PID: 2067 Comm: ln Not tainted 3.18.0-00587-gdfa582e #1
[ 70.332511] task: dd9c92c0 ti: dd822000 task.ti: dd822000
[ 70.333094] PC is at function_alloc_mtp_ptp+0xe/0x68
[ 70.333311] LR is at usb_get_function+0x11/0x1c
[ 70.333489] pc : [<c034ec12>] lr : [<c033cce9>] psr: 60070033
<..snip..>
[ 70.384111] 3fc0: bec14ae4 00000004 bec14c0a 00000053 00000004 b6f0422d 00000000 bec14adc
[ 70.384369] 3fe0: bec14af8 bec14a98 b6f071f3 b6e8977c 20070010 bec14c0d 00000000 00000000
[ 70.384832] [<c034ec12>] (function_alloc_mtp_ptp) from [<c033cce9>] (usb_get_function+0x11/0x1c)
[ 70.385146] [<c033cce9>] (usb_get_function) from [<c033da9b>] (config_usb_cfg_link+0x87/0xa8)
[ 70.385421] [<c033da9b>] (config_usb_cfg_link) from [<c011f417>] (configfs_symlink+0xb7/0x1c8)
[ 70.385696] [<c011f417>] (configfs_symlink) from [<c00dcd8d>] (vfs_symlink+0x85/0xc0)
[ 70.386010] [<c00dcd8d>] (vfs_symlink) from [<c00dce0b>] (SyS_symlinkat+0x43/0x70)
[ 70.386261] [<c00dce0b>] (SyS_symlinkat) from [<c000ce41>] (ret_fast_syscall+0x1/0x5c)
[ 70.386610] Code: eb04 4a0f 6e03 480f (e883) 0005
[ 70.387346] ---[ end trace 8dba7c552e02f8fa ]---
[ 70.387647] Kernel panic - not syncing: Fatal exception
[ 70.387980] ---[ end Kernel panic - not syncing: Fatal exception
-----------------------
Steps to reproduce the kernel panic:
mount -t configfs none /config
mkdir /config/usb_gadget/g1
cd /config/usb_gadget/g1
echo 0x18d1 > idVendor
echo 0x4e26 > idProduct
mkdir strings/0x409
echo 0123459876 > strings/0x409/serialnumber
echo Asus > strings/0x409/manufacturer
echo Nexus7 > strings/0x409/product
mkdir configs/c.1
mkdir configs/c.1/strings/0x409
echo "Conf 1" > configs/c.1/strings/0x409/configuration
echo 120 > configs/c.1/MaxPower
mkdir functions/ptp.ptp
ln -s functions/ptp.ptp configs/c.1/ptp.ptp
Also MTP and PTP are mutually exclusive functions
so make sure we have only one of it linked to a
configuration at a time. Otherwise it opens up
another set of bug(s?).
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
Added a device attribute to android_device to
determine USB_GADGET's state
Signed-off-by: Badhri Jagan Sridharan <Badhri@google.com>
Change-Id: I17f8903120df96bf2f4bf441940b53a87b818230
configfs_composite_unbind sets the gadget data
to null. Therefore, add check in disconnect
function to make sure that cdev is not NULL.
Prints a WARN message if the driver tries to
redundantly disconnect a gadget.
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
Change-Id: I248cb7175d0dd9a51c18053dd39475d8b3284f6d
This patch adds support to use Android accessory
gadget function through the
DECLARE_USB_FUNCTION_INIT interface.
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
Change-Id: Ib352752d5bc905fa1df9049b53eabf1294930db7
This patch adds support to use audio_source
gadget function through DECLARE_USB_FUNCTION_INIT
interface.
Signed-off-by: Badhri Jagan Sridharan <badhri@google.com>
Change-Id: I1fc6c9ea07105ae4eb785eebd3bb925bfdd8bc6b
Added create_function_device to create child
function devices for USB gadget functions.
Android UsbDeviceManager relies on communicating
to the devices created by the gadget functions
to implement functions such as audio_source.
Signed-off-by: Badhri Jagan Sridharan <Badhri@google.com>
Change-Id: I0df9ad86ac32d8cdacdea164e9fed49891b45fc2
While disabling ConfigFS Android gadget, android_disconnect() calls
kill_all_hid_devices(), if CONFIG_USB_CONFIGFS_F_ACC is enabled, to free
the registered HIDs without checking whether the USB accessory device
really exist or not. If USB accessory device doesn't exist then we run into
following kernel panic:
----8<----
[ 136.724761] Unable to handle kernel NULL pointer dereference at virtual address 00000064
[ 136.724809] pgd = c0204000
[ 136.731924] [00000064] *pgd=00000000
[ 136.737830] Internal error: Oops: 5 [#1] SMP ARM
[ 136.738108] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.18.0-rc4-00400-gf75300e-dirty #76
[ 136.742788] task: c0fb19d8 ti: c0fa4000 task.ti: c0fa4000
[ 136.750890] PC is at _raw_spin_lock_irqsave+0x24/0x60
[ 136.756246] LR is at kill_all_hid_devices+0x24/0x114
---->8----
This patch adds a test to check if USB Accessory device exists before freeing HIDs.
Change-Id: Ie229feaf0de3f4f7a151fcaa9a994e34e15ff73b
Signed-off-by: Amit Pundir <amit.pundir@linaro.org>
