Skip to contentSkip to conversation

: xl: don't free string literals

PATCH In review Gianni Tedesco 3 participants 6 messages Sep 8, 2010 – Sep 9, 2010

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 : xl: don't free string literals

The function init_dm_info() is initialising some strings from literals. This is bad juju because when the destructor is called we cannot know if the string literal was overridden with a strdup()'d value. Therefore strdup values in the initialiser then introduce and use the function libxlu_cfg_replace_string() which free's whatever is set before strdupping the new value on top of it. The rule for the new call should be clear due to const vs. non-const arguments - changing the behaviour of libxlu_cfg_get_string() would cause more complexity than it saves.

tools/libxl/libxlu_cfg.c +12 -1
diff -r e14a3c281982 tools/libxl/libxlu_cfg.c
@@ -151,7 +151,18 @@ int xlu_cfg_get_string(const XLU_Config 
     *value_r= set->values[0];
     return 0;
 }
-        
+ 
+int xlu_cfg_replace_string(const XLU_Config *cfg, const char *n,
+                           char **value_r) {
+    XLU_ConfigSetting *set;
+    int e;
+
+    e= find_atom(cfg,n,&set);  if (e) return e;
+    free(*value_r);
+    *value_r= strdup(set->values[0]);
+    return 0;
+}
+
 int xlu_cfg_get_long(const XLU_Config *cfg, const char *n,
                      long *value_r) {
     long l;
tools/libxl/libxlutil.h +1 -0
diff -r e14a3c281982 tools/libxl/libxlutil.h
@@ -46,6 +46,7 @@ void xlu_cfg_destroy(XLU_Config*);
  */
 
 int xlu_cfg_get_string(const XLU_Config*, const char *n, const char **value_r);
+int xlu_cfg_replace_string(const XLU_Config *cfg, const char *n, char **value_r); /* free/strdup version */
 int xlu_cfg_get_long(const XLU_Config*, const char *n, long *value_r);
 
 int xlu_cfg_get_list(const XLU_Config*, const char *n,
tools/libxl/xl_cmdimpl.c +11 -19
diff -r e14a3c281982 tools/libxl/xl_cmdimpl.c
@@ -297,7 +297,7 @@ static void init_dm_info(libxl_device_mo
     libxl_uuid_generate(&dm_info->uuid);
 
     dm_info->dom_name = c_info->name;
-    dm_info->device_model = "qemu-dm";
+    dm_info->device_model = strdup("qemu-dm");
     dm_info->videoram = b_info->video_memkb / 1024;
     dm_info->apic = b_info->u.hvm.apic;
     dm_info->vcpus = b_info->max_vcpus;
@@ -305,7 +305,7 @@ static void init_dm_info(libxl_device_mo
 
     dm_info->stdvga = 0;
     dm_info->vnc = 1;
-    dm_info->vnclisten = "127.0.0.1";
+    dm_info->vnclisten = strdup("127.0.0.1");
     dm_info->vncdisplay = 0;
     dm_info->vncunused = 1;
     dm_info->keymap = NULL;
@@ -313,7 +313,7 @@ static void init_dm_info(libxl_device_mo
     dm_info->opengl = 0;
     dm_info->nographic = 0;
     dm_info->serial = NULL;
-    dm_info->boot = "cda";
+    dm_info->boot = strdup("cda");
     dm_info->usb = 0;
     dm_info->usbdevice = NULL;
     dm_info->xen_platform_pci = 1;
@@ -1022,38 +1022,30 @@ skip_vfb:
         init_dm_info(dm_info, c_info, b_info);
 
         /* then process config related to dm */
-        if (!xlu_cfg_get_string (config, "device_model", &buf))
-            dm_info->device_model = strdup(buf);
+        !xlu_cfg_replace_string (config, "device_model", &dm_info->device_model);
         if (!xlu_cfg_get_long (config, "stdvga", &l))
             dm_info->stdvga = l;
         if (!xlu_cfg_get_long (config, "vnc", &l))
             dm_info->vnc = l;
-        if (!xlu_cfg_get_string (config, "vnclisten", &buf))
-            dm_info->vnclisten = strdup(buf);
-        if (!xlu_cfg_get_string (config, "vncpasswd", &buf))
-            dm_info->vncpasswd = strdup(buf);
+        xlu_cfg_replace_string (config, "vnclisten", &dm_info->vnclisten);
+        xlu_cfg_replace_string (config, "vncpasswd", &dm_info->vncpasswd);
         if (!xlu_cfg_get_long (config, "vncdisplay", &l))
             dm_info->vncdisplay = l;
         if (!xlu_cfg_get_long (config, "vncunused", &l))
             dm_info->vncunused = l;
-        if (!xlu_cfg_get_string (config, "keymap", &buf))
-            dm_info->keymap = strdup(buf);
+        xlu_cfg_replace_string (config, "keymap", &dm_info->keymap);
         if (!xlu_cfg_get_long (config, "sdl", &l))
             dm_info->sdl = l;
         if (!xlu_cfg_get_long (config, "opengl", &l))
             dm_info->opengl = l;
         if (!xlu_cfg_get_long (config, "nographic", &l))
             dm_info->nographic = l;
-        if (!xlu_cfg_get_string (config, "serial", &buf))
-            dm_info->serial = strdup(buf);
-        if (!xlu_cfg_get_string (config, "boot", &buf))
-            dm_info->boot = strdup(buf);
+        xlu_cfg_replace_string (config, "serial", &dm_info->serial);
+        xlu_cfg_replace_string (config, "boot", &dm_info->boot);
         if (!xlu_cfg_get_long (config, "usb", &l))
             dm_info->usb = l;
-        if (!xlu_cfg_get_string (config, "usbdevice", &buf))
-            dm_info->usbdevice = strdup(buf);
-        if (!xlu_cfg_get_string (config, "soundhw", &buf))
-            dm_info->soundhw = strdup(buf);
+        xlu_cfg_replace_string (config, "usbdevice", &dm_info->usbdevice);
+        xlu_cfg_replace_string (config, "soundhw", &dm_info->soundhw);
         if (!xlu_cfg_get_long (config, "xen_platform_pci", &l))
             dm_info->xen_platform_pci = l;
     }
#1 lore raw
Quoted 6 lines · Gianni Tedesco
         /* then process config related to dm */
tools/libxl/xl_cmdimpl.c +1 -2
-        if (!xlu_cfg_get_string (config, "device_model", &buf))
-            dm_info->device_model = strdup(buf);
+        !xlu_cfg_replace_string (config, "device_model",

&dm_info->device_model);

Stray ! at the start of the new line?

Ian.

#2 lore raw
Quoted 9 lines · Ian Campbell
Quoted 6 lines · Gianni Tedesco
         /* then process config related to dm */
tools/libxl/xl_cmdimpl.c +1 -2
-        if (!xlu_cfg_get_string (config, "device_model", &buf))
-            dm_info->device_model = strdup(buf);
+        !xlu_cfg_replace_string (config, "device_model",

&dm_info->device_model);

Stray ! at the start of the new line?

Indeed.

Jackson, I can re-post this or probably easier for you to fix up manually before committing?

Gianni

#3 lore raw
In reply to Gianni Tedesco
Quoted 10 lines · Gianni Tedesco

The function init_dm_info() is initialising some strings from literals. This is bad juju because when the destructor is called we cannot know if the string literal was overridden with a strdup()'d value. Therefore strdup values in the initialiser then introduce and use the function libxlu_cfg_replace_string() which free's whatever is set before strdupping the new value on top of it. The rule for the new call should be clear due to const vs. non-const arguments - changing the behaviour of libxlu_cfg_get_string() would cause more complexity than it saves.

About to bail for the day, but I guess this (untested) patch makes sense on top of this?

Avoids c_info->name being a string literal as well. Although I wonder if maybe domain configurations with no name could just be rejected as invalid? Having a whole bunch of domains called "test" doesn't seem likely to be what anyone wants...

Ian.

Subject: xl: use xlu_cfg_replace_string in a few more places.

tools/libxl/xl_cmdimpl.c +6 -12
diff -r d6026f6dcebf tools/libxl/xl_cmdimpl.c
@@ -606,10 +606,8 @@ static void parse_config_data(const char
     if (!xlu_cfg_get_long (config, "hap", &l))
         c_info->hap = l;
 
-    if (!xlu_cfg_get_string (config, "name", &buf))
-        c_info->name = strdup(buf);
-    else
-        c_info->name = "test";
+    if (xlu_cfg_replace_string (config, "name", &c_info->name))
+        c_info->name = strdup("test");
 
     if (!xlu_cfg_get_string (config, "uuid", &buf) ) {
         if ( libxl_uuid_from_string(&c_info->uuid, buf) ) {
@@ -695,8 +693,7 @@ static void parse_config_data(const char
     if (!xlu_cfg_get_long (config, "videoram", &l))
         b_info->video_memkb = l * 1024;
 
-    if (!xlu_cfg_get_string (config, "kernel", &buf))
-        b_info->kernel.path = strdup(buf);
+    xlu_cfg_replace_string (config, "kernel", &b_info->kernel.path);
 
     if (c_info->hvm == 1) {
         if (!xlu_cfg_get_long (config, "pae", &l))
@@ -734,10 +731,8 @@ static void parse_config_data(const char
             exit(1);
         }
 
-        if (!xlu_cfg_get_string (config, "bootloader", &buf))
-            b_info->u.pv.bootloader = strdup(buf);
-        if (!xlu_cfg_get_string (config, "bootloader_args", &buf))
-            b_info->u.pv.bootloader_args = strdup(buf);
+        xlu_cfg_replace_string (config, "bootloader", &b_info->u.pv.bootloader);
+        xlu_cfg_replace_string (config, "bootloader_args", &b_info->u.pv.bootloader_args);
 
         if (!b_info->u.pv.bootloader && !b_info->kernel.path) {
             fprintf(stderr, "Neither kernel nor bootloader specified\n");
@@ -745,8 +740,7 @@ static void parse_config_data(const char
         }
 
         b_info->u.pv.cmdline = cmdline;
-        if (!xlu_cfg_get_string (config, "ramdisk", &buf))
-            b_info->u.pv.ramdisk.path = strdup(buf);
+        xlu_cfg_replace_string (config, "ramdisk", &b_info->u.pv.ramdisk.path);
     }
 
     if (!xlu_cfg_get_list (config, "disk", &vbds, 0)) {
#4 lore raw
Quoted 10 lines · Ian Campbell
Quoted 9 lines · Gianni Tedesco

The function init_dm_info() is initialising some strings from literals. This is bad juju because when the destructor is called we cannot know if the string literal was overridden with a strdup()'d value. Therefore strdup values in the initialiser then introduce and use the function libxlu_cfg_replace_string() which free's whatever is set before strdupping the new value on top of it. The rule for the new call should be clear due to const vs. non-const arguments - changing the behaviour of libxlu_cfg_get_string() would cause more complexity than it saves.

Quoted 4 lines
Quoted 1 line

About to bail for the day, but I guess this (untested) patch makes sense on top of this?

Yes, very much so

Quoted 4 lines

Avoids c_info->name being a string literal as well. Although I wonder if maybe domain configurations with no name could just be rejected as invalid? Having a whole bunch of domains called "test" doesn't seem likely to be what anyone wants...

Yeah, that looks suspect to me too.

Gianni

#5 lore raw
In reply to Ian Campbell

Ian Campbell writes ("Re: [Xen-devel] [PATCH,v2]: xl: don't free string literals"):

Quoted 2 lines

About to bail for the day, but I guess this (untested) patch makes sense on top of this?

...

Quoted 1 line

Subject: xl: use xlu_cfg_replace_string in a few more places.

Yes, I have applied it.

Quoted 4 lines

Avoids c_info->name being a string literal as well. Although I wonder if maybe domain configurations with no name could just be rejected as invalid? Having a whole bunch of domains called "test" doesn't seem likely to be what anyone wants...

Yes. Really we should be rejecting duplicate names.

Ian.