Develop Bugs

[Q] root_plug

Not sure if this means anything, just throwing it out there, but i was going through file for froyo on the evo, original file name "supersonic-2.6.32.15-g746f4f0.tar" and found a file named root_plug-notepad, now i'm only guessing, but is this how they are blocking us from being able to root our device, or am i just guessing outta my arse.
on the notepad the full content is:
/*
* Root Plug sample LSM module
*
* Originally written for a Linux Journal.
*
* Copyright (C) 2002 Greg Kroah-Hartman
*
* Prevents any programs running with egid == 0 if a specific USB device
* is not present in the system. Yes, it can be gotten around, but is a
* nice starting point for people to play with, and learn the LSM
* interface.
*
* If you want to turn this into something with a semblance of security,
* you need to hook the task_* functions also.
*
* See for more information
* about this code.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation, version 2 of the
* License.
*/
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/security.h>
#include <linux/usb.h>
#include <linux/moduleparam.h>
/* default is a generic type of usb to serial converter */
static int vendor_id = 0x0557;
static int product_id = 0x2008;
module_param(vendor_id, uint, 0400);
module_param(product_id, uint, 0400);
/* should we print out debug messages */
static int debug = 0;
module_param(debug, bool, 0600);
#define MY_NAME "root_plug"
#define root_dbg(fmt, arg...) \
do { \
if (debug) \
printk(KERN_DEBUG "%s: %s: " fmt , \
MY_NAME , __func__ , \
## arg); \
} while (0)
static int rootplug_bprm_check_security (struct linux_binprm *bprm)
{
struct usb_device *dev;
root_dbg("file %s, e_uid = %d, e_gid = %d\n",
bprm->filename, bprm->cred->euid, bprm->cred->egid);
if (bprm->cred->egid == 0) {
dev = usb_find_device(vendor_id, product_id);
if (!dev) {
root_dbg("e_gid = 0, and device not found, "
"task not allowed to run...\n");
return -EPERM;
}
usb_put_dev(dev);
}
return 0;
}
static struct security_operations rootplug_security_ops = {
.bprm_check_security = rootplug_bprm_check_security,
};
static int __init rootplug_init (void)
{
/* register ourselves with the security framework */
if (register_security (&rootplug_security_ops)) {
printk (KERN_INFO
"Failure registering Root Plug module with the kernel\n");
return -EINVAL;
}
printk (KERN_INFO "Root Plug module initialized, "
"vendor_id = %4.4x, product id = %4.4x\n", vendor_id, product_id);
return 0;
}
security_initcall (rootplug_init);

nobody, nothing, someone has to have some sort of input, lol.

pubbs.net/200910/kernel/36760-patchrfc-security-remove-rootplug.html

Uncle jimmy says hello

[Q] Ermmm...is this good or bad?

In a previous forum post, we indicated that we would look into this request and, unfortunately we are not able to support video codecs for ARMv6 OMX IL libs for Qualcomm's QDSP5 on MSM7x27. The architecture of MSM7x27 cannot support OpenMax video codecs for ARMv6 OMX IL on QDSP5 running on Android Ice Cream Sandwich (ICS).
Click to expand...
Click to collapse
What does this mean? Does it relate to our phone?

Yes, it does relate to our phones. The Wildfire S has a Qualcomm MSM7227 chipset. Apparently, OpenMax is a set of open source libraries which are useful for video decoding, picture capturing, video recording and more on the hardware chip (The DSP Processor [GPU], in this case, the Adreno 200) directly and not on the CPU. In our case (From the message you posted), these libraries cannot work with the built-in hardware dsp processor so they can't use it to provide hardware decoding of videos on ICS.
Found the post at https://developer.qualcomm.com/forum/qdevnet-forums/multimedia-optimization-qdsp/13248
I'll try looking through it to see if I can understand any more details.
Okay. It says that gingerbread used the OMX IL libraries but the version which was provided was a lower version which is not completely compatible with ICS.
These drivers are not open source so they cannot be worked on by developers. They have to be released by Qualcomm directly. But according to the post, it would seem that the new OMX IL libraries which are required for ICS are not compatible with the armv6 instruction set and are only compatible with armv7 onwards.

Do we know the version that we have libraries for and the version we need ?
Okay. It says that gingerbread used the OMX IL libraries but the version which was provided was a lower version which is not completely compatible with ICS.
Do we know what version of the OMX IL libraries we have for gb and what version we need for ICS ? Thinking we might be able to write a compatibility layer to sit between them, if there isn't too much difference.

omx libraries.
scottrix said:
Okay. It says that gingerbread used the OMX IL libraries but the version which was provided was a lower version which is not completely compatible with ICS.
Do we know what version of the OMX IL libraries we have for gb and what version we need for ICS ? Thinking we might be able to write a compatibility layer to sit between them, if there isn't too much difference.
Click to expand...
Click to collapse
Don't know the versions yet, but the binaries are here....
vendor/htc/marvel/proprietary/lib/libOmxH264Dec.so
vendor/htc/marvel/proprietary/lib/libOmxMpeg4Dec.so
vendor/htc/marvel/proprietary/lib/libOmxVidEnc.so

scottrix said:
Don't know the versions yet, but the binaries are here....
vendor/htc/marvel/proprietary/lib/libOmxH264Dec.so
vendor/htc/marvel/proprietary/lib/libOmxMpeg4Dec.so
vendor/htc/marvel/proprietary/lib/libOmxVidEnc.so
Click to expand...
Click to collapse
More information in these posts:
http://www.modaco.com/topic/354020-zte-n880e-first-zte-ics-update/page__st__20
http://forum.xda-developers.com/archive/index.php/t-1588599-p-2.html
http://newsandeducationblog.blogspot.com/2012/05/android-developers-video-videoview.html
http://forum.xda-developers.com/archive/index.php/t-1664335.html
---------- Post added at 09:01 PM ---------- Previous post was at 08:53 PM ----------
scottrix said:
More information in these posts:
http://www.modaco.com/topic/354020-zte-n880e-first-zte-ics-update/page__st__20
http://forum.xda-developers.com/archive/index.php/t-1588599-p-2.html
http://newsandeducationblog.blogspot.com/2012/05/android-developers-video-videoview.html
http://forum.xda-developers.com/archive/index.php/t-1664335.html
Click to expand...
Click to collapse
For get_config the debug output:
W/QCvdec ( 2751): get_config: unknown param 117440527
117440527 is 0x700000F, which from frameworks/base/include/media/stagefright/openmax/OMX_Index.h
/* Image & Video common Configurations */
OMX_IndexCommonStartUnused = 0x07000000,
OMX_IndexParamCommonDeblocking, /**< reference: OMX_PARAM_DEBLOCKINGTYPE */
OMX_IndexParamCommonSensorMode, /**< reference: OMX_PARAM_SENSORMODETYPE */
OMX_IndexParamCommonInterleave, /**< reference: OMX_PARAM_INTERLEAVETYPE */
OMX_IndexConfigCommonColorFormatConversion, /**< reference: OMX_CONFIG_COLORCONVERSIONTYPE */
OMX_IndexConfigCommonScale, /**< reference: OMX_CONFIG_SCALEFACTORTYPE */
OMX_IndexConfigCommonImageFilter, /**< reference: OMX_CONFIG_IMAGEFILTERTYPE */
OMX_IndexConfigCommonColorEnhancement, /**< reference: OMX_CONFIG_COLORENHANCEMENTTYPE */
OMX_IndexConfigCommonColorKey, /**< reference: OMX_CONFIG_COLORKEYTYPE */
OMX_IndexConfigCommonColorBlend, /**< reference: OMX_CONFIG_COLORBLENDTYPE */
OMX_IndexConfigCommonFrameStabilisation,/**< reference: OMX_CONFIG_FRAMESTABTYPE */
OMX_IndexConfigCommonRotate, /**< reference: OMX_CONFIG_ROTATIONTYPE */
OMX_IndexConfigCommonMirror, /**< reference: OMX_CONFIG_MIRRORTYPE */
OMX_IndexConfigCommonOutputPosition, /**< reference: OMX_CONFIG_POINTTYPE */
OMX_IndexConfigCommonInputCrop, /**< reference: OMX_CONFIG_RECTTYPE */
OMX_IndexConfigCommonOutputCrop, /**< reference: OMX_CONFIG_RECTTYPE */
OMX_IndexConfigCommonDigitalZoom, /**< reference: OMX_CONFIG_SCALEFACTORTYPE */
Is OMX_IndexConfigCommonOutputCrop, so the hardware mpeg4 decoder doesn't understand this index. Going from the source code that calls this, if the call fails then it does something else and carries on
./frameworks/base/media/libstagefright/ACodec.cpp:
if (mOMX->getConfig(
mNode, OMX_IndexConfigCommonOutputCrop,
&rect, sizeof(rect)) != OK) {
rect.nLeft = 0;
rect.nTop = 0;
rect.nWidth = videoDef->nFrameWidth;
rect.nHeight = videoDef->nFrameHeight;
}
CHECK_GE(rect.nLeft, 0);
CHECK_GE(rect.nTop, 0);
CHECK_GE(rect.nWidth, 0u);
CHECK_GE(rect.nHeight, 0u);
CHECK_LE(rect.nLeft + rect.nWidth - 1, videoDef->nFrameWidth);
CHECK_LE(rect.nTop + rect.nHeight - 1, videoDef->nFrameHeight);
./frameworks/base/media/libstagefright/OMXCodec.cpp:
status_t err =
mOMX->getConfig(
mNode, OMX_IndexConfigCommonOutputCrop,
&rect, sizeof(rect));
CODEC_LOGI(
"video dimensions are %ld x %ld",
video_def->nFrameWidth, video_def->nFrameHeight);
if (err == OK) {
#ifdef SAMSUNG_CODEC_SUPPORT
/* Hack GetConfig */
rect.nLeft = 0;
rect.nTop = 0;
rect.nWidth = video_def->nFrameWidth;
rect.nHeight = video_def->nFrameHeight;
#endif
CHECK_GE(rect.nLeft, 0);
CHECK_GE(rect.nTop, 0);
CHECK_GE(rect.nWidth, 0u);
CHECK_GE(rect.nHeight, 0u);
CHECK_LE(rect.nLeft + rect.nWidth - 1, video_def->nFrameWidth);
CHECK_LE(rect.nTop + rect.nHeight - 1, video_def->nFrameHeight);
mOutputFormat->setRect(
kKeyCropRect,
rect.nLeft,
rect.nTop,
rect.nLeft + rect.nWidth - 1,
rect.nTop + rect.nHeight - 1);
CODEC_LOGI(
"Crop rect is %ld x %ld @ (%ld, %ld)",
rect.nWidth, rect.nHeight, rect.nLeft, rect.nTop);
} else {
mOutputFormat->setRect(
kKeyCropRect,
0, 0,
video_def->nFrameWidth - 1,
video_def->nFrameHeight - 1);
}
I think we can safely ignore this failure for now.
---------- Post added at 09:30 PM ---------- Previous post was at 09:01 PM ----------
get_extension_index might be a little more tricky. The file vendor/qcom/opensource/omx/mm-video/vidc/vdec/src/omx_vdec.cpp is a video decoder and the code for the get_extension_index is:
OMX_ERRORTYPE omx_vdec::get_extension_index(OMX_IN OMX_HANDLETYPE hComp,
OMX_IN OMX_STRING paramName,
OMX_OUT OMX_INDEXTYPE* indexType)
{
if(m_state == OMX_StateInvalid)
{
DEBUG_PRINT_ERROR("Get Extension Index in Invalid State\n");
return OMX_ErrorInvalidState;
}
else if (!strncmp(paramName, "OMX.QCOM.index.param.video.SyncFrameDecodingMode",sizeof("OMX.QCOM.index.param.video.SyncFrameDecodingMode") - 1)) {
*indexType = (OMX_INDEXTYPE)OMX_QcomIndexParamVideoSyncFrameDecodingMode;
}
#ifdef MAX_RES_1080P
else if (!strncmp(paramName, "OMX.QCOM.index.param.IndexExtraData",sizeof("OMX.QCOM.index.param.IndexExtraData") - 1))
{
*indexType = (OMX_INDEXTYPE)OMX_QcomIndexParamIndexExtraDataType;
}
#endif
#if defined (_ANDROID_HONEYCOMB_) || defined (_ANDROID_ICS_)
else if(!strncmp(paramName,"OMX.google.android.index.enableAndroidNativeBuffers", sizeof("OMX.google.android.index.enableAndroidNativeBuffers") - 1)) {
*indexType = (OMX_INDEXTYPE)OMX_GoogleAndroidIndexEnableAndroidNativeBuffers;
}
else if(!strncmp(paramName,"OMX.google.android.index.useAndroidNativeBuffer2", sizeof("OMX.google.android.index.enableAndroidNativeBuffer2") - 1)) {
*indexType = (OMX_INDEXTYPE)OMX_GoogleAndroidIndexUseAndroidNativeBuffer2;
}
else if(!strncmp(paramName,"OMX.google.android.index.useAndroidNativeBuffer", sizeof("OMX.google.android.index.enableAndroidNativeBuffer") - 1)) {
DEBUG_PRINT_ERROR("Extension: %s is supported\n", paramName);
*indexType = (OMX_INDEXTYPE)OMX_GoogleAndroidIndexUseAndroidNativeBuffer;
}
else if(!strncmp(paramName,"OMX.google.android.index.getAndroidNativeBufferUsage", sizeof("OMX.google.android.index.getAndroidNativeBufferUsage") - 1)) {
*indexType = (OMX_INDEXTYPE)OMX_GoogleAndroidIndexGetAndroidNativeBufferUsage;
}
#endif
else {
DEBUG_PRINT_ERROR("Extension: %s not implemented\n", paramName);
return OMX_ErrorNotImplemented;
}
return OMX_ErrorNone;
}
In the debug output the message "OMX_GetExtensionIndex failed" comes from frameworks/base/media/libstagefright/omx/OMXNodeInstance.cpp:
status_t OMXNodeInstance::enableGraphicBuffers(
OMX_U32 portIndex, OMX_BOOL enable) {
Mutex::Autolock autoLock(mLock);
OMX_INDEXTYPE index;
OMX_ERRORTYPE err = OMX_GetExtensionIndex(
mHandle,
const_cast<OMX_STRING>("OMX.google.android.index.enableAndroidNativeBuffers"),
&index);
if (err != OMX_ErrorNone) {
LOGE("OMX_GetExtensionIndex failed");
return StatusFromOMXError(err);
}
OMX_VERSIONTYPE ver;
ver.s.nVersionMajor = 1;
ver.s.nVersionMinor = 0;
ver.s.nRevision = 0;
ver.s.nStep = 0;
EnableAndroidNativeBuffersParams params = {
sizeof(EnableAndroidNativeBuffersParams), ver, portIndex, enable,
};
err = OMX_SetParameter(mHandle, index, &params);
if (err != OMX_ErrorNone) {
LOGE("OMX_EnableAndroidNativeBuffers failed with error %d (0x%08x)",
err, err);
return UNKNOWN_ERROR;
}
return OK;
}
So the OMX.google.android.index.enableAndroidNativeBuffers extension is the one we are missing, from the code snippet above this, we can see an
#if defined (_ANDROID_HONEYCOMB_) || defined (_ANDROID_ICS_)
#endif
around the check for this extension, so it has been added since Gingerbread, a diff with the cm7 source tells us that the enableGraphicsBuffers has also been added since gingerbread. The header file /frameworks/base/include/media/stagefright/HardwareAPI.h contains the following:
// A pointer to this struct is passed to the OMX_SetParameter when the extension
// index for the 'OMX.google.android.index.enableAndroidNativeBuffers' extension
// is given.
//
// When Android native buffer use is disabled for a port (the default state),
// the OMX node should operate as normal, and expect UseBuffer calls to set its
// buffers. This is the mode that will be used when CPU access to the buffer is
// required.
//
// When Android native buffer use has been enabled for a given port, the video
// color format for the port is to be interpreted as an Android pixel format
// rather than an OMX color format. The node should then expect to receive
// UseAndroidNativeBuffer calls (via OMX_SetParameter) rather than UseBuffer
// calls for that port.
struct EnableAndroidNativeBuffersParams {
OMX_U32 nSize;
OMX_VERSIONTYPE nVersion;
OMX_U32 nPortIndex;
OMX_BOOL enable;
};
I doubt the gb codec that we do have has support for the Android pixel format.
Question is can we change the stagefright sources to support both, i.e. if the extension OMX.google.android.index.enableAndroidNativeBuffers isn't supported then do what was done in gb. I'm sure it is possible and it might still be possible to get a codec wrapper working, however, it will take a lot more digging in the code.

[UTIL][PB]★(¯ `•.MS ® Windows Phone - for Flash.Store.Bin[Ext/Build/Info/More].•´¯)★

[UTIL][PB]★(¯ `•.MS ® Windows Phone - for Flash.Store.Bin[Ext/Build/Info/More].•´¯)★
Now We have tool for Ext/Build/Mod./and More Flash.Store.bin
imgtool - Modyfication Image Store, extract/build/repartition/and more... for system
structure store partition, and general part sections image Windows Phone.
Support All Image Flash.Store.Bin Devices:
-- Toshiba/Dell/HTC/Samsung/Nokia/Etc...
Support Other Image System Devices:
-- bin|dio|xml|raw
Info - Usage:
Code:
ImgTool [...options...] inFile [outFile]
Info - Options:
Code:
Options:
-inFormat bin|dio|xml|raw Input file format
-inCert certificateFile Certificate for input file
-inBase address Raw image base address
-outFormat bin|xml|raw|secbin Output file format
-outHash sum|sha1 Output file hash
-outPvk privateKeyFile Private key for output hash
-outPvkPwd password Password for private key
-outCert certificateFile Certificate for output file
-ui Enable crypto user interface
-sectorSize bytes Output file sector size
-sectorsPerRecord sectors Output file sectors in record
-recordSize Output file maximal record size
-ram2store [...type...] Converts ram to storage image
-extendLast Set last partition size as unlimited
-extractPublicKey Extract public key from input certificate
Platform Info © 2012 By UPE-Product
All Rights Reserved

Support Example Format Store File - BOOT_DOWNLOAD_BIN_FORMAT_STORE:
Code:
//
// Copyright (c) Microsoft Corporation. All rights reserved.
//
//
// Use of this sample source code is subject to the terms of the Microsoft
// license agreement under which you licensed this sample source code. If
// you did not accept the terms of the license agreement, you are not
// authorized to use this sample source code. For the terms of the license,
// please see the license agreement between you and Microsoft or, if applicable,
// see the LICENSE.RTF on your install media or the root of your tools installation.
// THE SAMPLE SOURCE CODE IS PROVIDED "AS IS", WITH NO WARRANTIES OR INDEMNITIES.
//
#ifndef __BOOT_DOWNLOAD_BIN_FORMAT_H
#define __BOOT_DOWNLOAD_BIN_FORMAT_H
#include "bootTypes.h"
#ifdef __cplusplus
extern "C" {
#endif
#pragma pack(push, 1)
//------------------------------------------------------------------------------
#define BOOT_BIN_SIGNATURE_RAM "B000FF\x0A"
#define BOOT_BIN_SIGNATURE_SIGNED_RAM "S000FF\x0A"
#define BOOT_BIN_SIGNATURE_RAW "N000FF\x0A"
#define BOOT_BIN_SIGNATURE_SIGNED_RAW "R000FF\x0A"
#define BOOT_BIN_SIGNATURE_STORE "D000FF\x0A"
#define BOOT_BIN_SIGNATURE_JUMP "J00000\x0A"
//------------------------------------------------------------------------------
typedef struct BootBinFormatRamHeader_t {
uint32_t start;
uint32_t size;
} BootBinFormatRamHeader_t;
typedef struct BootBinFormatRamRecordHeader_t {
uint32_t address;
uint32_t length;
uint32_t checksum;
} BootBinFormatRamRecordHeader_t;
//------------------------------------------------------------------------------
typedef struct BootBinFormatSignature_t {
uint8_t signature[7];
} BootBinFormatSignature_t;
enum BootBinFormatFlags_e {
BOOT_BIN_FORMAT_FLAG_CLEAN = (1 << 0)
};
//------------------------------------------------------------------------------
enum BootBinFormatHashType_e {
BOOT_BIN_FORMAT_HASH_SUM = 0,
BOOT_BIN_FORMAT_HASH_SHA1 = 1,
BOOT_BIN_FORMAT_HASH_SHA256 = 2,
BOOT_BIN_FORMAT_HASH_OEM = 0x8000
// OEM can define new hashing algorithms in their own platform-specific
// BSPs. The definitions must start at BOOT_BIN_FORMAT_HASH_OEM, e.g.
// BOOT_BIN_FORMAT_HASH_ALG1 = BOOT_BIN_FORMAT_HASH_OEM + 0,
// BOOT_BIN_FORMAT_HASH_ALG2 = BOOT_BIN_FORMAT_HASH_OEM + 1
};
//------------------------------------------------------------------------------
typedef struct BootBinFormatStoreHeader_t {
uint32_t flags;
uint32_t sectorSize;
uint32_t sectors;
uint32_t segments;
uint32_t hashType;
uint32_t hashSize;
uint32_t hashInfoSize;
} BootBinFormatStoreHeader_t, *BootBinFormatStoreHeader;
typedef struct BootBinFormatStoreSha1Info_t {
uint32_t publicKeySize;
uint32_t seedSize;
/*
uint8_t publicKey[publicKeySize];
uint8_t seed[seedSize];
*/
} BootBinFormatStoreSha1Info_t;
enum BootBinFormatStoreSegmentType_e {
BOOT_BIN_FORMAT_STORE_SEGMENT_BINARY = 1,
BOOT_BIN_FORMAT_STORE_SEGMENT_RESERVED = 2,
BOOT_BIN_FORMAT_STORE_SEGMENT_PARTITION = 3
};
typedef struct BootBinFormatStoreSegmentHeader_t {
uint32_t type;
uint32_t sectors;
uint32_t infoSize;
} BootBinFormatStoreSegmentHeader_t, *BootBinFormatStoreSegmentHeader;
typedef struct BootBinFormatStoreSegmentBinaryInfo_t {
uint8_t index;
} BootBinFormatStoreSegmentBinaryInfo_t;
typedef struct BootBinFormatStoreSegmentReservedInfo_t {
char name[8];
} BootBinFormatStoreSegmentReservedInfo_t;
typedef struct BootBinFormatStoreSegmentPartitionInfo_t {
uint8_t fileSystem;
uint8_t index;
} BootBinFormatStoreSegmentPartitionInfo_t;
typedef struct BootBinFormatStoreRecordHeader_t {
uint32_t segment;
uint32_t sector;
uint32_t sectors;
} BootBinFormatStoreRecordHeader_t;
/*
This is pseudo-C definition of store format
typedef struct BootBinFormatStore_t {
uint8_t signature[7];
BootBinFormatStoreHeader_t header;
uint8_t hashInfo[header.hashInfoSize];
struct {
BootBinFormatStoreSegmentHeader_t segmentHeader;
union {
BootBinFormatStoreSegmentBinaryInfo_t binaryInfo;
BootBinFormatStoreSegmentReservedInfo_t reservedInfo;
BootBinFormatStoreSegmentPartitionInfo_t partitionInfo;
};
} segment[header.segments];
uint8_t hash[header.hashSize]; // Header hash/checksum
struct {
BootBinFormatStoreRecordHeader_t recordHeader;
uint8_t recordData[recordHeader.sectors * header.sectorSize];
uint8_t recordHash[header.hashSize];
} record[];
} BootBinFormatStore_t;
*/
//------------------------------------------------------------------------------
#pragma pack(pop)
#ifdef __cplusplus
}
#endif
#endif // __BOOT_DOWNLOAD_BIN_FORMAT_H

Unblocking thread safe queue - check code, please (and use, if right for you)

Hi friends.
I must solve big problem to finish HaRET (and other native applications) managed wrapper and UI (i. e. for example managed Silverlight UI controlled from unmanaged code), safe WP7 FTP client, interprocess cooperation on WP7, etc.
I know it is a basic issue, which is probably solved many time in operating systems (windows messages etc.) and many other examples.
But, I could find no any source code for queue, which is:
1. safe for more writers (threads from more processes) and for one reader,
2. unblocking for all writers,
3. unblocking for reader.
I found only very much unblocking ringbuffers or blocking queues, no one unblocking unlimited queue. Please, send corrections if I am bad looking.
Then, I tried code it...
First I want use std::queue with Thread Local Storage identifications (one subqueue for all writer and one abqueue for reader). But, TLS is very limited on CE based systems and I am not sure, if it solve interprocess safety too.
Finally, I tried to write the following code. It is designed for interthread queue only still, but interprocess can be added by safe named filemapping probably. Tell your opinion, it will be the basis for many complex applications here. For many years I have not worked with low-level synchronisation, I could make a fundamental mistake. So far I've missed templates and any wrapper, it's just the basic idea, usable in both C/C++ with very little changes:
PHP:
// This queue is unblocking and thread safe, when:
// 1. more writers are allowed,
// 2. one reader is allowed only.
// Motto: Pointers queue is both-directional, but frontal direction is changed by writers, backward direction by reader only!
#ifndef InterlockedExchangePointer
#define InterlockedExchangePointer(Target, Value) ((PVOID)InterlockedExchange((PLONG)(Target), (LONG)(Value)))
#endif
#ifndef InterlockedCompareExchangePointer
#define InterlockedCompareExchangePointer(Destination, ExChange, Comperand) ((PVOID)InterlockedCompareExchange((PLONG)(Destination), (LONG)(ExChange), (LONG)(Comperand)))
#endif
typedef struct SQItem
{
SQItem * pNext;
SQItem * pPrev;
void * pValue;
} * PItem;
SQItem EmptyItem = {NULL, NULL, NULL};
PItem pFront = &EmptyItem;
PItem pBack = &EmptyItem;
// Empty queue contains one item with both directions pointers nulled.
bool push_front(void * pNewValue) // nonblocking push, but theoretically forever cycle is possible for low priority thread! Practically improbable.
{
if (!pNewValue)
{
return false;
}
PItem pNewFront = new SQItem;
if (pNewFront)
{
PItem pProbablyFront;
InterlockedExchangePointer(&pProbablyFront, pFront);
pNewFront->pNext = pProbablyFront;
pNewFront->pPrev = NULL;
pNewFront->pValue = pNewValue;
while (InterlockedCompareExchangePointer(&pFront, pNewFront, pProbablyFront) != pProbablyFront)
{ // Another writer changed front pointer, active writer must try it again.
InterlockedExchangePointer(&pProbablyFront, pFront);
pNewFront->pNext = pProbablyFront;
}
return true; // Success allways.
}
return false; // Insufficient memory.
}
void * pop_back() // one reader allowed only!!!
{
PItem pBackward;
InterlockedExchangePointer(&pBackward, pFront); // Safe copy of actual front pointer
if (pBackward == pBack) // Queue is empty in time of interlocked attempt.
{
return NULL;
}
if (!pBack->pPrev)
{ // If backward queue is not continued here, reconstruct it from safe front pointer.
while (pBackward != pBack && !pBackward->pNext->pPrev)
{
pBackward->pNext->pPrev = pBackward;
pBackward = pBackward->pNext;
}
}
// One reader can safe read from back and manipulate with backward queue.
PItem pDelete = pBack;
pBack->pPrev->pNext = NULL;
pBack = pBack->pPrev;
delete pDelete;
return pBack->pValue;
/* // This is not necessary, but it is more clean:
void * pResult = pBack->pValue;
pBack->pValue = NULL;
return pResult;
*/
}
I am not sure, if heap allocation is not more "blocking" operation due to swapping and mm priorities, then object/signal waiting. What do you mean about it?
Mirror principle can be used to OneToMore queue making. Thanks unblocking, MoreToOne and OneToMore can be immeditelly piped to MoreToMore one.
But, when CE systems do not support two coupled pointers interlocked functions, there is no possibility to identify allowed reader safely, then this has not great significance.
As well, MoreToOne is much more needed now.

Template version
Using (simple example, the same way will be used between native threads and managed Silverlight UI):
Code:
#include "..\Templates\STQueue.h"
STQueue<tstring> qMessages(L"");
Code:
... // Called by [B]any thread[/B]:
void SafeMessage(tstring sMessage)
{
qMessages.push_front(sMessage);
PostMessage(hUnsafe, WM_FTP_MESSAGE, 0 , 0);
}
...
Code:
... // In UI thread:
LRESULT OnMsgMessage(HWND hWnd, WPARAM wParam, LPARAM lParam)
{
bool bChange = false;
tstring szMessage = L"";
while ((szMessage = qMessages.pop_back()) != L"")
{
szMessages += szMessage;
szMessages += L"\r\n";
bChange = true;
}
if (bChange)
{
SetDlgItemText(hMainDlg, IDC_EDITMessage, szMessages.c_str());
}
return TRUE;
}
...
STQueue.h:
PHP:
template <class _Type> class STQItem
{
public:
STQItem<_Type> * pNext;
STQItem<_Type> * pPrev;
_Type Value;
STQItem(_Type & rV, STQItem<_Type> * pN = NULL, STQItem<_Type> * pP = NULL);
};
template <class _Type> class STQueue
{
protected:
STQItem<_Type> * pFront;
STQItem<_Type> * pBack;
_Type EmptyValue;
public:
STQueue(_Type EmptyV);
~STQueue();
bool push_front(const _Type & rNewValue);
const _Type & pop_back();
bool empty();
};
template <class _Type> STQItem<_Type>::STQItem(_Type & rV, STQItem<_Type> * pN, STQItem<_Type> * pP)
{
Value = rV;
pNext = pN;
pPrev = pP;
}
template <class _Type> STQueue<_Type>::STQueue(_Type EmptyV)
{
EmptyValue = EmptyV;
STQItem<_Type> * EmtyItem = new STQItem<_Type>(EmptyValue);
pFront = EmtyItem;
pBack = EmtyItem;
}
template <class _Type> STQueue<_Type>::~STQueue()
{
while (pop_back() != EmptyValue)
{
}
delete pBack;
}
template <class _Type> bool STQueue<_Type>::push_front(const _Type & rNewValue)
{ // nonblocking push, but theoretically forever cycle is possible for low priority thread!
STQItem<_Type> * pNewFront = new STQItem<_Type>(EmptyValue);
if (pNewFront)
{
STQItem<_Type> * pProbablyFront;
InterlockedExchangePointer(&pProbablyFront, pFront);
pNewFront->pNext = pProbablyFront;
pNewFront->pPrev = NULL;
pNewFront->Value = rNewValue;
while (InterlockedCompareExchangePointer(&pFront, pNewFront, pProbablyFront) != pProbablyFront)
{ // Another writer changed front pointer, active writer must try it again.
InterlockedExchangePointer(&pProbablyFront, pFront);
pNewFront->pNext = pProbablyFront;
}
return true; // Success allways.
}
return false; // Insufficient memory.
}
template <class _Type> const _Type & STQueue<_Type>::pop_back()
{ // one reader allowed only!!!
STQItem<_Type> * pBackward;
InterlockedExchangePointer(&pBackward, pFront); // Safe copy of actual front pointer
if (pBackward == pBack) // Queue is empty in time of interlocked attempt.
{
return EmptyValue;
}
if (!pBack->pPrev)
{ // If backward queue is not beginned, reconstruct it from safe front pointer.
while (pBackward != pBack && !pBackward->pNext->pPrev)
{
pBackward->pNext->pPrev = pBackward;
pBackward = pBackward->pNext;
}
}
// One reader can safe read from back and manipulate with backward queue.
STQItem<_Type> * pDelete = pBack;
pBack->pPrev->pNext = NULL;
pBack = pBack->pPrev;
delete pDelete;
return pBack->Value;
/* // This is not necessary, but it is more clean:
void * pResult = pBack->pValue;
pBack->pValue = NULL;
return pResult;
*/
}
template <class _Type> bool STQueue<_Type>::empty()
{
if(last==NULL)
return TRUE;
else
return FALSE;
}

[Q] How to get the PERF_COUNT_HW_INSTRUCTION event with perf_event on Nexus 7 2013?

I'm trying to get hardware performance counter events on Nexus 7 2013.
I've tested on ASOP 4.3, 4.4, CyanogenMod 11, google original kernel 3.4, and codeaurora msm-3.4 kernel.
I already tested that CPU_CYCLE(PERF_COUNT_HW_CPU_CYCLES) and L2 Cache related events are working.
However, I couldn't get an executed instrunction event (PERF_COUNT_HW_INSTRUCTION) which is one of the default HPC events.
It always returns zero!
I'm stuck for 2 weeks.:crying:
Please somebody help me!
Thanks in advance.
My test code is as follow.
Code:
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <sys/ioctl.h>
#include <linux/perf_event.h>
#include <asm/unistd.h>
static long
perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
int cpu, int group_fd, unsigned long flags)
{
int ret;
ret = syscall(__NR_perf_event_open, hw_event, pid, cpu,
group_fd, flags);
return ret;
}
int
main(int argc, char **argv)
{
struct perf_event_attr pe;
long long count;
int fd;
int i=0;
memset(&pe, 0, sizeof(struct perf_event_attr));
pe.type = PERF_TYPE_HARDWARE;
pe.size = sizeof(struct perf_event_attr);
pe.config = PERF_COUNT_HW_INSTRUCTIONS; //PERF_COUNT_HW_CPU_CYCLES;
pe.disabled = 1;
pe.exclude_kernel = 1;
pe.exclude_hv = 1;
fd = perf_event_open(&pe, 0, 0, -1, 0);
if (fd == -1) {
fprintf(stderr, "Error opening leader %llx\n", pe.config);
exit(EXIT_FAILURE);
}
ioctl(fd, PERF_EVENT_IOC_RESET, 0);
ioctl(fd, PERF_EVENT_IOC_ENABLE, 0);
for (i=0; i< 10000; i++)
printf("%d = %d * %d\n", i*(i+1), i, i+1);
ioctl(fd, PERF_EVENT_IOC_DISABLE, 0);
read(fd, &count, sizeof(long long));
printf("%lld events\n", count);
close(fd);
}

Same here
I've also had no luck I'm actually not able to get any hardware counters (they are all zero or errno 2). I only got some software counters like the clock. Did you need cygenomod to get cycles/cache counters working? I figured this should work even without custom builds. Did you need to run as root while querying them?