Adds hyperscan support to pm operator

Author: xiangwang1

src/Makefile.am

@@ -245,6 +245,7 @@ UTILS = \
 	utils/decode.cc \
 	utils/geo_lookup.cc \
 	utils/https_client.cc \
+	utils/hyperscan.cc \
 	utils/ip_tree.cc \
 	utils/md5.cc \
 	utils/msc_tree.cc \

src/operators/pm.cc

@@ -25,11 +25,6 @@
 #include <list>
 #include <memory>
 
-#ifdef WITH_HS
-#include <hs.h>
-#endif
-
-
 #include "src/operators/operator.h"
 #ifndef WITH_HS
 #include "src/utils/acmp.h"
@@ -41,17 +36,18 @@ namespace operators {
 
 Pm::~Pm() {
 #ifdef WITH_HS
+    m_hs = NULL;
 #else
     acmp_node_t *root = m_p->root_node;
 
     cleanup(root);
 
     free(m_p);
     m_p = NULL;
+#endif
 #ifdef MODSEC_MUTEX_ON_PM
     pthread_mutex_destroy(&m_lock);
 #endif
-#endif
 }
 
 #ifndef WITH_HS
@@ -95,22 +91,28 @@ void Pm::postOrderTraversal(acmp_btree_node_t *node) {
 
 bool Pm::evaluate(Transaction *transaction, RuleWithActions *rule,
     const std::string &input, std::shared_ptr<RuleMessage> ruleMessage) {
+    int rc = 0;
+    const char *match = NULL;
 #ifdef WITH_HS
-    return 0;
+#ifdef MODSEC_MUTEX_ON_PM
+    pthread_mutex_lock(&m_lock);
+#endif
+    rc = m_hs->search(input.c_str(), input.length(), &match);
+#ifdef MODSEC_MUTEX_ON_PM
+    pthread_mutex_unlock(&m_lock);
+#endif
 #else
-    int rc;
     ACMPT pt;
     pt.parser = m_p;
     pt.ptr = NULL;
-    const char *match = NULL;
 #ifdef MODSEC_MUTEX_ON_PM
     pthread_mutex_lock(&m_lock);
 #endif
     rc = acmp_process_quick(&pt, &match, input.c_str(), input.length());
 #ifdef MODSEC_MUTEX_ON_PM
     pthread_mutex_unlock(&m_lock);
 #endif
-
+#endif
     if (rc >= 0 && transaction) {
         std::string match_(match?match:"");
         logOffset(ruleMessage, rc - match_.size() + 1, match_.size());
@@ -125,16 +127,138 @@ bool Pm::evaluate(Transaction *transaction, RuleWithActions *rule,
     }
 
     return rc >= 0;
+}
+
+static
+char *parse_pm_content(const char *op_parm, unsigned short int op_len, const char **error_msg)  {
+    char *parm = NULL;
+    char *content;
+    unsigned short int offset = 0;
+//    char converted = 0;
+    int i, x;
+    unsigned char bin = 0, esc = 0, bin_offset = 0;
+    unsigned char c;
+    unsigned char bin_parm[3] = { 0 };
+    char *processed = NULL;
+
+    content = strdup(op_parm);
+
+    if (content == NULL) {
+        *error_msg = std::string("Error allocating memory for pattern matching content.").c_str();
+        return NULL;
+    }
+
+    while (offset < op_len && (content[offset] == ' ' || content[offset] == '\t')) {
+        offset++;
+    };
+
+    op_len = strlen(content);
+
+    if (content[offset] == '\"' && content[op_len-1] == '\"') {
+        parm = strdup(content + offset + 1);
+        if (parm  == NULL) {
+            *error_msg = std::string("Error allocating memory for pattern matching content.").c_str();
+            free(content);
+            content = NULL;
+            return NULL;
+        }
+        parm[op_len - offset - 2] = '\0';
+    } else {
+        parm = strdup(content + offset);
+        if (parm == NULL) {
+            free(content);
+            content = NULL;
+            *error_msg = std::string("Error allocating memory for pattern matching content.").c_str();
+            return NULL;
+        }
+    }
+
+    free(content);
+    content = NULL;
+
+    op_len = strlen(parm);
+
+    if (op_len == 0)   {
+        *error_msg = "Content length is 0.";
+        free(parm);
+        return NULL;
+    }
+
+    for (i = 0, x = 0; i < op_len; i++) {
+        if (parm[i] == '|') {
+            if (bin) {
+                bin = 0;
+            } else {
+                bin = 1;
+            }
+        } else if(!esc && parm[i] == '\\') {
+            esc = 1;
+        } else {
+            if (bin) {
+                if (parm[i] == 0 || parm[i] == 1 || parm[i] == 2 ||
+                    parm[i] == 3 || parm[i] == 4 || parm[i] == 5 ||
+                    parm[i] == 6 || parm[i] == 7 || parm[i] == 8 ||
+                    parm[i] == 9 ||
+                    parm[i] == 'A' || parm[i] == 'a' ||
+                    parm[i] == 'B' || parm[i] == 'b' ||
+                    parm[i] == 'C' || parm[i] == 'c' ||
+                    parm[i] == 'D' || parm[i] == 'd' ||
+                    parm[i] == 'E' || parm[i] == 'e' ||
+                    parm[i] == 'F' || parm[i] == 'f')
+                {
+                    bin_parm[bin_offset] = (char)parm[i];
+                    bin_offset++;
+                    if (bin_offset == 2) {
+                        c = strtol((char *)bin_parm, (char **) NULL, 16) & 0xFF;
+                        bin_offset = 0;
+                        parm[x] = c;
+                        x++;
+                        //converted = 1;
+                    }
+                } else if (parm[i] == ' ') {
+                }
+            } else if (esc) {
+                if (parm[i] == ':' ||
+                        parm[i] == ';' ||
+                        parm[i] == '\\' ||
+                        parm[i] == '\"')
+                {
+                    parm[x] = parm[i];
+                    x++;
+                } else {
+                    *error_msg = std::string("Unsupported escape sequence.").c_str();
+                    free(parm);
+                    return NULL;
+                }
+                esc = 0;
+                //converted = 1;
+            } else {
+                parm[x] = parm[i];
+                x++;
+            }
+        }
+    }
+
+#if 0
+    if (converted) {
+        op_len = x;
+    }
 #endif
 
-    return 0;
-}
+    //processed = memcpy(processed, parm, op_len);
+    processed = strdup(parm);
+    free(parm);
+    parm = NULL;
 
+    if (processed == NULL) {
+        *error_msg = std::string("Error allocating memory for pattern matching content.").c_str();
+        return NULL;
+    }
+
+    return processed;
+}
 
 bool Pm::init(const std::string &file, std::string *error) {
-#ifdef WITH_HS
-    fprintf(stdout, "Sopport for HS is on the way: %s\n", hs_version());
-#else
     std::vector<std::string> vec;
     std::istringstream *iss;
     const char *err = NULL;
@@ -154,20 +278,32 @@ bool Pm::init(const std::string &file, std::string *error) {
         back_inserter(vec));
 
     for (auto &a : vec) {
+#ifdef WITH_HS
+        m_hs->addPattern(a.c_str(), a.length());
+    }
+    if (m_hs->compile(error) == false) {
+        if (content) {
+            free(content);
+            content = NULL;
+        }
+        delete iss;
+        return false;
+    }
+#else
         acmp_add_pattern(m_p, a.c_str(), NULL, NULL, a.length());
     }
 
     while (m_p->is_failtree_done == 0) {
         acmp_prepare(m_p);
     }
+#endif
 
     if (content) {
         free(content);
         content = NULL;
     }
 
     delete iss;
-#endif
     return true;
 }
 

src/operators/pm.h

@@ -22,8 +22,11 @@
 #include <utility>
 
 #include "src/operators/operator.h"
+#ifdef WITH_HS
+#include "src/utils/hyperscan.h"
+#else
 #include "src/utils/acmp.h"
-
+#endif
 
 namespace modsecurity {
 namespace operators {
@@ -34,15 +37,13 @@ class Pm : public Operator {
     /** @ingroup ModSecurity_Operator */
     explicit Pm(std::unique_ptr<RunTimeString> param)
         : Operator("Pm", std::move(param)) {
-#ifdef WITH_HS
-#else
+#ifndef WITH_HS
         m_p = acmp_create(0);
 #endif
     }
     explicit Pm(const std::string &n, std::unique_ptr<RunTimeString> param)
         : Operator(n, std::move(param)) {
-#ifdef WITH_HS
-#else
+#ifndef WITH_HS
         m_p = acmp_create(0);
 #endif
     }
@@ -59,16 +60,17 @@ class Pm : public Operator {
 #endif
 
  protected:
-#ifndef WITH_HS
+#ifdef WITH_HS
+    std::shared_ptr<Utils::HyperscanPm> m_hs =
+        std::make_shared<Utils::HyperscanPm>();
+#else
     ACMP *m_p;
 #endif
 
  private:
-#ifndef WITH_HS
 #ifdef MODSEC_MUTEX_ON_PM
     pthread_mutex_t m_lock;
 #endif
-#endif
 
 };
 

src/operators/pm_from_file.cc

@@ -69,13 +69,18 @@ bool PmFromFile::init(const std::string &config, std::string *error) {
     for (std::string line; std::getline(*iss, line); ) {
         if (isComment(line) == false) {
 #ifdef WITH_HS
+            m_hs->addPattern(line.c_str(), line.length());
+        }
+    }
+    if (m_hs->compile(error) == false) {
+        delete iss;
+        return false;
+    }
 #else
             acmp_add_pattern(m_p, line.c_str(), NULL, NULL, line.length());
-#endif
-	}
+	    }
     }
 
-#ifndef WITH_HS
     while (m_p->is_failtree_done == 0) {
         acmp_prepare(m_p);
     }