Implement a new section to attach LAST_KMSG to incident report
This section simply gzip a large file and stores result in GZippedFileProto

This greatly improves the size, before gzip, the last kmsg size ~500KB,
after gzip the proto size is ~60KB.

Bug: 73354384
Test: atest incidentd_test and manual on device test
Change-Id: I9bfc2cf07384487671edbffb5f0bd8495608fea6
diff --git a/cmds/incidentd/src/FdBuffer.cpp b/cmds/incidentd/src/FdBuffer.cpp
index db60794..64da677 100644
--- a/cmds/incidentd/src/FdBuffer.cpp
+++ b/cmds/incidentd/src/FdBuffer.cpp
@@ -76,6 +76,7 @@
                         return -errno;
                     }
                 } else if (amt == 0) {
+                    VLOG("Reached EOF of fd=%d", fd);
                     break;
                 }
                 mBuffer.wp()->move(amt);
@@ -156,10 +157,10 @@
                 if (!(errno == EAGAIN || errno == EWOULDBLOCK)) {
                     VLOG("Fail to read fd %d: %s", fd, strerror(errno));
                     return -errno;
-                }                   // otherwise just continue
-            } else if (amt == 0) {  // reach EOF so don't have to poll pfds[0].
-                ::close(pfds[0].fd);
-                pfds[0].fd = -1;
+                }  // otherwise just continue
+            } else if (amt == 0) {
+                VLOG("Reached EOF of input file %d", fd);
+                pfds[0].fd = -1;  // reach EOF so don't have to poll pfds[0].
             } else {
                 rpos += amt;
                 cirSize += amt;
@@ -187,6 +188,7 @@
 
         // if buffer is empty and fd is closed, close write fd.
         if (cirSize == 0 && pfds[0].fd == -1 && pfds[1].fd != -1) {
+            VLOG("Close write pipe %d", toFd);
             ::close(pfds[1].fd);
             pfds[1].fd = -1;
         }
@@ -207,6 +209,7 @@
                 return -errno;
             }  // otherwise just continue
         } else if (amt == 0) {
+            VLOG("Reached EOF of fromFd %d", fromFd);
             break;
         } else {
             mBuffer.wp()->move(amt);
diff --git a/cmds/incidentd/src/FdBuffer.h b/cmds/incidentd/src/FdBuffer.h
index 5bfa093..66a3de1 100644
--- a/cmds/incidentd/src/FdBuffer.h
+++ b/cmds/incidentd/src/FdBuffer.h
@@ -26,7 +26,7 @@
 using namespace std;
 
 /**
- * Reads a file into a buffer, and then writes that data to an FdSet.
+ * Reads data from fd into a buffer, fd must be closed explicitly.
  */
 class FdBuffer {
 public:
@@ -83,6 +83,11 @@
      */
     EncodedBuffer::iterator data() const;
 
+    /**
+     * Return the internal buffer, don't call unless you are familiar with EncodedBuffer.
+     */
+    EncodedBuffer* getInternalBuffer() { return &mBuffer; }
+
 private:
     EncodedBuffer mBuffer;
     int64_t mStartTime;
diff --git a/cmds/incidentd/src/Section.cpp b/cmds/incidentd/src/Section.cpp
index 64eae3a..334d77c 100644
--- a/cmds/incidentd/src/Section.cpp
+++ b/cmds/incidentd/src/Section.cpp
@@ -18,12 +18,8 @@
 
 #include "Section.h"
 
-#include <errno.h>
-#include <sys/prctl.h>
-#include <unistd.h>
 #include <wait.h>
 
-#include <memory>
 #include <mutex>
 
 #include <android-base/file.h>
@@ -37,6 +33,7 @@
 #include "FdBuffer.h"
 #include "Privacy.h"
 #include "PrivacyBuffer.h"
+#include "frameworks/base/core/proto/android/os/data.proto.h"
 #include "frameworks/base/core/proto/android/util/log.proto.h"
 #include "incidentd_util.h"
 
@@ -52,31 +49,11 @@
 const int WAIT_MAX = 5;
 const struct timespec WAIT_INTERVAL_NS = {0, 200 * 1000 * 1000};
 const char INCIDENT_HELPER[] = "/system/bin/incident_helper";
+const char GZIP[] = "/system/bin/gzip";
 
-static pid_t fork_execute_incident_helper(const int id, const char* name, Fpipe& p2cPipe,
-                                          Fpipe& c2pPipe) {
+static pid_t fork_execute_incident_helper(const int id, Fpipe* p2cPipe, Fpipe* c2pPipe) {
     const char* ihArgs[]{INCIDENT_HELPER, "-s", String8::format("%d", id).string(), NULL};
-    // fork used in multithreaded environment, avoid adding unnecessary code in child process
-    pid_t pid = fork();
-    if (pid == 0) {
-        if (TEMP_FAILURE_RETRY(dup2(p2cPipe.readFd(), STDIN_FILENO)) != 0 || !p2cPipe.close() ||
-            TEMP_FAILURE_RETRY(dup2(c2pPipe.writeFd(), STDOUT_FILENO)) != 1 || !c2pPipe.close()) {
-            ALOGW("%s can't setup stdin and stdout for incident helper", name);
-            _exit(EXIT_FAILURE);
-        }
-
-        /* make sure the child dies when incidentd dies */
-        prctl(PR_SET_PDEATHSIG, SIGKILL);
-
-        execv(INCIDENT_HELPER, const_cast<char**>(ihArgs));
-
-        ALOGW("%s failed in incident helper process: %s", name, strerror(errno));
-        _exit(EXIT_FAILURE);  // always exits with failure if any
-    }
-    // close the fds used in incident helper
-    close(p2cPipe.readFd());
-    close(c2pPipe.writeFd());
-    return pid;
+    return fork_execute_cmd(INCIDENT_HELPER, const_cast<char**>(ihArgs), p2cPipe, c2pPipe);
 }
 
 // ================================================================================
@@ -254,10 +231,12 @@
     return NO_ERROR;
 }
 // ================================================================================
+static inline bool isSysfs(const char* filename) { return strncmp(filename, "/sys/", 5) == 0; }
+
 FileSection::FileSection(int id, const char* filename, const int64_t timeoutMs)
     : Section(id, timeoutMs), mFilename(filename) {
     name = filename;
-    mIsSysfs = strncmp(filename, "/sys/", 5) == 0;
+    mIsSysfs = isSysfs(filename);
 }
 
 FileSection::~FileSection() {}
@@ -280,7 +259,7 @@
         return -errno;
     }
 
-    pid_t pid = fork_execute_incident_helper(this->id, this->name.string(), p2cPipe, c2pPipe);
+    pid_t pid = fork_execute_incident_helper(this->id, &p2cPipe, &c2pPipe);
     if (pid == -1) {
         ALOGW("FileSection '%s' failed to fork", this->name.string());
         return -errno;
@@ -289,6 +268,8 @@
     // parent process
     status_t readStatus = buffer.readProcessedDataInStream(fd, p2cPipe.writeFd(), c2pPipe.readFd(),
                                                            this->timeoutMs, mIsSysfs);
+    close(fd);  // close the fd anyway.
+
     if (readStatus != NO_ERROR || buffer.timedOut()) {
         ALOGW("FileSection '%s' failed to read data from incident helper: %s, timedout: %s",
               this->name.string(), strerror(-readStatus), buffer.timedOut() ? "true" : "false");
@@ -313,7 +294,99 @@
 
     return NO_ERROR;
 }
+// ================================================================================
+GZipSection::GZipSection(int id, const char* filename, ...) : Section(id) {
+    name = "gzip ";
+    name += filename;
+    va_list args;
+    va_start(args, filename);
+    mFilenames = varargs(filename, args);
+    va_end(args);
+}
 
+GZipSection::~GZipSection() {}
+
+status_t GZipSection::Execute(ReportRequestSet* requests) const {
+    // Reads the files in order, use the first available one.
+    int index = 0;
+    int fd = -1;
+    while (mFilenames[index] != NULL) {
+        fd = open(mFilenames[index], O_RDONLY | O_CLOEXEC);
+        if (fd != -1) {
+            break;
+        }
+        ALOGW("GZipSection failed to open file %s", mFilenames[index]);
+        index++;  // look at the next file.
+    }
+    VLOG("GZipSection is using file %s, fd=%d", mFilenames[index], fd);
+    if (fd == -1) return -1;
+
+    FdBuffer buffer;
+    Fpipe p2cPipe;
+    Fpipe c2pPipe;
+    // initiate pipes to pass data to/from gzip
+    if (!p2cPipe.init() || !c2pPipe.init()) {
+        ALOGW("GZipSection '%s' failed to setup pipes", this->name.string());
+        return -errno;
+    }
+
+    const char* gzipArgs[]{GZIP, NULL};
+    pid_t pid = fork_execute_cmd(GZIP, const_cast<char**>(gzipArgs), &p2cPipe, &c2pPipe);
+    if (pid == -1) {
+        ALOGW("GZipSection '%s' failed to fork", this->name.string());
+        return -errno;
+    }
+    // parent process
+
+    // construct Fdbuffer to output GZippedfileProto, the reason to do this instead of using
+    // ProtoOutputStream is to avoid allocation of another buffer inside ProtoOutputStream.
+    EncodedBuffer* internalBuffer = buffer.getInternalBuffer();
+    internalBuffer->writeHeader((uint32_t)GZippedFileProto::FILENAME, WIRE_TYPE_LENGTH_DELIMITED);
+    String8 usedFile(mFilenames[index]);
+    internalBuffer->writeRawVarint32(usedFile.size());
+    for (size_t i = 0; i < usedFile.size(); i++) {
+        internalBuffer->writeRawByte(mFilenames[index][i]);
+    }
+    internalBuffer->writeHeader((uint32_t)GZippedFileProto::GZIPPED_DATA,
+                                WIRE_TYPE_LENGTH_DELIMITED);
+    size_t editPos = internalBuffer->wp()->pos();
+    internalBuffer->wp()->move(8);  // reserve 8 bytes for the varint of the data size.
+    size_t dataBeginAt = internalBuffer->wp()->pos();
+    VLOG("GZipSection '%s' editPos=%zd, dataBeginAt=%zd", this->name.string(), editPos,
+         dataBeginAt);
+
+    status_t readStatus = buffer.readProcessedDataInStream(
+            fd, p2cPipe.writeFd(), c2pPipe.readFd(), this->timeoutMs, isSysfs(mFilenames[index]));
+    close(fd);  // close the fd anyway.
+
+    if (readStatus != NO_ERROR || buffer.timedOut()) {
+        ALOGW("GZipSection '%s' failed to read data from gzip: %s, timedout: %s",
+              this->name.string(), strerror(-readStatus), buffer.timedOut() ? "true" : "false");
+        kill_child(pid);
+        return readStatus;
+    }
+
+    status_t gzipStatus = wait_child(pid);
+    if (gzipStatus != NO_ERROR) {
+        ALOGW("GZipSection '%s' abnormal child process: %s", this->name.string(),
+              strerror(-gzipStatus));
+        return gzipStatus;
+    }
+    // Revisit the actual size from gzip result and edit the internal buffer accordingly.
+    size_t dataSize = buffer.size() - dataBeginAt;
+    internalBuffer->wp()->rewind()->move(editPos);
+    internalBuffer->writeRawVarint32(dataSize);
+    internalBuffer->copy(dataBeginAt, dataSize);
+    VLOG("GZipSection '%s' wrote %zd bytes in %d ms, dataSize=%zd", this->name.string(),
+         buffer.size(), (int)buffer.durationMs(), dataSize);
+    status_t err = write_report_requests(this->id, buffer, requests);
+    if (err != NO_ERROR) {
+        ALOGW("GZipSection '%s' failed writing: %s", this->name.string(), strerror(-err));
+        return err;
+    }
+
+    return NO_ERROR;
+}
 // ================================================================================
 struct WorkerThreadData : public virtual RefBase {
     const WorkerThreadSection* section;
@@ -457,42 +530,20 @@
 }
 
 // ================================================================================
-void CommandSection::init(const char* command, va_list args) {
-    va_list copied_args;
-    int numOfArgs = 0;
-
-    va_copy(copied_args, args);
-    while (va_arg(copied_args, const char*) != NULL) {
-        numOfArgs++;
-    }
-    va_end(copied_args);
-
-    // allocate extra 1 for command and 1 for NULL terminator
-    mCommand = (const char**)malloc(sizeof(const char*) * (numOfArgs + 2));
-
-    mCommand[0] = command;
-    name = command;
-    for (int i = 0; i < numOfArgs; i++) {
-        const char* arg = va_arg(args, const char*);
-        mCommand[i + 1] = arg;
-        name += " ";
-        name += arg;
-    }
-    mCommand[numOfArgs + 1] = NULL;
-}
-
 CommandSection::CommandSection(int id, const int64_t timeoutMs, const char* command, ...)
     : Section(id, timeoutMs) {
+    name = command;
     va_list args;
     va_start(args, command);
-    init(command, args);
+    mCommand = varargs(command, args);
     va_end(args);
 }
 
 CommandSection::CommandSection(int id, const char* command, ...) : Section(id) {
+    name = command;
     va_list args;
     va_start(args, command);
-    init(command, args);
+    mCommand = varargs(command, args);
     va_end(args);
 }
 
@@ -527,7 +578,7 @@
               strerror(errno));
         _exit(err);  // exit with command error code
     }
-    pid_t ihPid = fork_execute_incident_helper(this->id, this->name.string(), cmdPipe, ihPipe);
+    pid_t ihPid = fork_execute_incident_helper(this->id, &cmdPipe, &ihPipe);
     if (ihPid == -1) {
         ALOGW("CommandSection '%s' failed to fork", this->name.string());
         return -errno;
@@ -544,8 +595,7 @@
     }
 
     // TODO: wait for command here has one trade-off: the failed status of command won't be detected
-    // until
-    //       buffer timeout, but it has advatage on starting the data stream earlier.
+    // until buffer timeout, but it has advatage on starting the data stream earlier.
     status_t cmdStatus = wait_child(cmdPid);
     status_t ihStatus = wait_child(ihPid);
     if (cmdStatus != NO_ERROR || ihStatus != NO_ERROR) {
diff --git a/cmds/incidentd/src/Section.h b/cmds/incidentd/src/Section.h
index d644681..8294be1 100644
--- a/cmds/incidentd/src/Section.h
+++ b/cmds/incidentd/src/Section.h
@@ -84,6 +84,21 @@
 };
 
 /**
+ * Section that reads in a file and gzips the content.
+ */
+class GZipSection : public Section {
+public:
+    GZipSection(int id, const char* filename, ...);
+    virtual ~GZipSection();
+
+    virtual status_t Execute(ReportRequestSet* requests) const;
+
+private:
+    // It looks up the content from multiple files and stops when the first one is available.
+    const char** mFilenames;
+};
+
+/**
  * Base class for sections that call a command that might need a timeout.
  */
 class WorkerThreadSection : public Section {
@@ -111,8 +126,6 @@
 
 private:
     const char** mCommand;
-
-    void init(const char* command, va_list args);
 };
 
 /**
diff --git a/cmds/incidentd/src/incidentd_util.cpp b/cmds/incidentd/src/incidentd_util.cpp
index 2415860..fc7cec9 100644
--- a/cmds/incidentd/src/incidentd_util.cpp
+++ b/cmds/incidentd/src/incidentd_util.cpp
@@ -13,8 +13,13 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
+#define DEBUG false
+#include "Log.h"
+
 #include "incidentd_util.h"
 
+#include <sys/prctl.h>
+
 #include "section_list.h"
 
 const Privacy* get_privacy_of_section(int id) {
@@ -50,4 +55,49 @@
 
 int Fpipe::readFd() const { return mRead.get(); }
 
-int Fpipe::writeFd() const { return mWrite.get(); }
\ No newline at end of file
+int Fpipe::writeFd() const { return mWrite.get(); }
+
+pid_t fork_execute_cmd(const char* cmd, char* const argv[], Fpipe* input, Fpipe* output) {
+    // fork used in multithreaded environment, avoid adding unnecessary code in child process
+    pid_t pid = fork();
+    if (pid == 0) {
+        if (TEMP_FAILURE_RETRY(dup2(input->readFd(), STDIN_FILENO)) < 0 || !input->close() ||
+            TEMP_FAILURE_RETRY(dup2(output->writeFd(), STDOUT_FILENO)) < 0 || !output->close()) {
+            ALOGW("Can't setup stdin and stdout for command %s", cmd);
+            _exit(EXIT_FAILURE);
+        }
+
+        /* make sure the child dies when incidentd dies */
+        prctl(PR_SET_PDEATHSIG, SIGKILL);
+
+        execv(cmd, argv);
+
+        ALOGW("%s failed in the child process: %s", cmd, strerror(errno));
+        _exit(EXIT_FAILURE);  // always exits with failure if any
+    }
+    // close the fds used in child process.
+    close(input->readFd());
+    close(output->writeFd());
+    return pid;
+}
+// ================================================================================
+const char** varargs(const char* first, va_list rest) {
+    va_list copied_rest;
+    int numOfArgs = 1;  // first is already count.
+
+    va_copy(copied_rest, rest);
+    while (va_arg(copied_rest, const char*) != NULL) {
+        numOfArgs++;
+    }
+    va_end(copied_rest);
+
+    // allocate extra 1 for NULL terminator
+    const char** ret = (const char**)malloc(sizeof(const char*) * (numOfArgs + 1));
+    ret[0] = first;
+    for (int i = 0; i < numOfArgs; i++) {
+        const char* arg = va_arg(rest, const char*);
+        ret[i + 1] = arg;
+    }
+    ret[numOfArgs + 1] = NULL;
+    return ret;
+}
diff --git a/cmds/incidentd/src/incidentd_util.h b/cmds/incidentd/src/incidentd_util.h
index 09aa040..db7ec82 100644
--- a/cmds/incidentd/src/incidentd_util.h
+++ b/cmds/incidentd/src/incidentd_util.h
@@ -20,12 +20,20 @@
 
 #include <android-base/unique_fd.h>
 
+#include <stdarg.h>
+
 #include "Privacy.h"
 
 using namespace android::base;
 
+/**
+ * Looks up Privacy of a section in the auto-gen PRIVACY_POLICY_LIST;
+ */
 const Privacy* get_privacy_of_section(int id);
 
+/**
+ * This class wraps android::base::Pipe.
+ */
 class Fpipe {
 public:
     Fpipe();
@@ -41,4 +49,15 @@
     unique_fd mWrite;
 };
 
+/**
+ * Forks and exec a command with two pipes, one connects stdin for input,
+ * one connects stdout for output. It returns the pid of the child.
+ */
+pid_t fork_execute_cmd(const char* cmd, char* const argv[], Fpipe* input, Fpipe* output);
+
+/**
+ * Grabs varargs from stack and stores them in heap with NULL-terminated array.
+ */
+const char** varargs(const char* first, va_list rest);
+
 #endif  // INCIDENTD_UTIL_H
\ No newline at end of file