[PATCH] qga: Use gethostname() instead of g_get_host_name()

[PATCH] qga: Use gethostname() instead of g_get_host_name()

[Michal Privoznik]

Problem with g_get_host_name() is that on the first call it saves
the hostname into a global variable and from then on, every
subsequent call returns the saved hostname. Even if the hostname
changes. This doesn't play nicely with guest agent, because if
the hostname is acquired before the guest is set up (e.g. on the
first boot, or before DHCP) we will report old, invalid hostname.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1845127

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
---
 qga/commands.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 52 insertions(+), 4 deletions(-)

diff --git a/qga/commands.c b/qga/commands.c
index efc8b90281..ce3c2041a6 100644
--- a/qga/commands.c
+++ b/qga/commands.c
@@ -512,14 +512,62 @@ int ga_parse_whence(GuestFileWhence *whence, Error **errp)
     return -1;
 }
 
+#ifndef HOST_NAME_MAX
+# ifdef _POSIX_HOST_NAME_MAX
+#  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
+# else
+#  define HOST_NAME_MAX 255
+# endif
+#endif
+
 GuestHostName *qmp_guest_get_host_name(Error **errp)
 {
     GuestHostName *result = NULL;
-    gchar const *hostname = g_get_host_name();
-    if (hostname != NULL) {
-        result = g_new0(GuestHostName, 1);
-        result->host_name = g_strdup(hostname);
+    g_autofree char *hostname = NULL;
+
+    /*
+     * We want to avoid using g_get_host_name() because that
+     * caches the result and we wouldn't reflect changes in the
+     * host name.
+     */
+
+#ifndef G_OS_WIN32
+    long len = -1;
+
+#ifdef _SC_HOST_NAME_MAX
+    len = sysconf(_SC_HOST_NAME_MAX);
+#endif /* _SC_HOST_NAME_MAX */
+
+    if (len < 0) {
+        len = HOST_NAME_MAX;
     }
+
+    hostname = g_malloc0(len + 1);
+
+    if (gethostname(hostname, len) < 0) {
+        return NULL;
+    }
+
+#else /* G_OS_WIN32 */
+
+    wchar_t tmp[MAX_COMPUTERNAME_LENGTH + 1];
+    DWORD size = G_N_ELEMENTS(tmp);
+
+    if (GetComputerNameW(tmp, &size) != 0) {
+        /*
+         * Indeed, on Windows retval of zero means failure
+         * and nonzero means success.
+         */
+        hostname = g_utf16_to_utf8(tmp, size, NULL, NULL, NULL);
+    }
+#endif /* G_OS_WIN32 */
+
+    if (!hostname) {
+        hostname = g_strdup("localhost");
+    }
+
+    result = g_new0(GuestHostName, 1);
+    result->host_name = g_steal_pointer(&hostname);
     return result;
 }
 
-- 
2.26.2

Re: [PATCH] qga: Use gethostname() instead of g_get_host_name()

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 2916 bytes --]

Hi

On Tue, Jun 16, 2020 at 12:35 PM Michal Privoznik <mprivozn@redhat.com>
wrote:

> Problem with g_get_host_name() is that on the first call it saves
> the hostname into a global variable and from then on, every
> subsequent call returns the saved hostname. Even if the hostname
> changes. This doesn't play nicely with guest agent, because if
> the hostname is acquired before the guest is set up (e.g. on the
> first boot, or before DHCP) we will report old, invalid hostname.
>
> Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1845127


lgtm,
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>

(fwiw, I opened a glib issue:
https://gitlab.gnome.org/GNOME/glib/-/issues/2130)


>
> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
> ---
>  qga/commands.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 52 insertions(+), 4 deletions(-)
>
> diff --git a/qga/commands.c b/qga/commands.c
> index efc8b90281..ce3c2041a6 100644
> --- a/qga/commands.c
> +++ b/qga/commands.c
> @@ -512,14 +512,62 @@ int ga_parse_whence(GuestFileWhence *whence, Error
> **errp)
>      return -1;
>  }
>
> +#ifndef HOST_NAME_MAX
> +# ifdef _POSIX_HOST_NAME_MAX
> +#  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
> +# else
> +#  define HOST_NAME_MAX 255
> +# endif
> +#endif
> +
>  GuestHostName *qmp_guest_get_host_name(Error **errp)
>  {
>      GuestHostName *result = NULL;
> -    gchar const *hostname = g_get_host_name();
> -    if (hostname != NULL) {
> -        result = g_new0(GuestHostName, 1);
> -        result->host_name = g_strdup(hostname);
> +    g_autofree char *hostname = NULL;
> +
> +    /*
> +     * We want to avoid using g_get_host_name() because that
> +     * caches the result and we wouldn't reflect changes in the
> +     * host name.
> +     */
> +
> +#ifndef G_OS_WIN32
> +    long len = -1;
> +
> +#ifdef _SC_HOST_NAME_MAX
> +    len = sysconf(_SC_HOST_NAME_MAX);
> +#endif /* _SC_HOST_NAME_MAX */
> +
> +    if (len < 0) {
> +        len = HOST_NAME_MAX;
>      }
> +
> +    hostname = g_malloc0(len + 1);
> +
> +    if (gethostname(hostname, len) < 0) {
> +        return NULL;
> +    }
> +
> +#else /* G_OS_WIN32 */
> +
> +    wchar_t tmp[MAX_COMPUTERNAME_LENGTH + 1];
> +    DWORD size = G_N_ELEMENTS(tmp);
> +
> +    if (GetComputerNameW(tmp, &size) != 0) {
> +        /*
> +         * Indeed, on Windows retval of zero means failure
> +         * and nonzero means success.
> +         */
> +        hostname = g_utf16_to_utf8(tmp, size, NULL, NULL, NULL);
> +    }
> +#endif /* G_OS_WIN32 */
> +
> +    if (!hostname) {
> +        hostname = g_strdup("localhost");
> +    }
> +
> +    result = g_new0(GuestHostName, 1);
> +    result->host_name = g_steal_pointer(&hostname);
>      return result;
>  }
>
> --
> 2.26.2
>
>
>

-- 
Marc-André Lureau

[-- Attachment #2: Type: text/html, Size: 4088 bytes --]

Re: [PATCH] qga: Use gethostname() instead of g_get_host_name()

[Richard Henderson]

On 6/16/20 1:34 AM, Michal Privoznik wrote:
> +#ifndef G_OS_WIN32

Nit: positive tests are easier to reason with and extend than negative tests.
I would reverse these two blocks and use a positive test for windows.

Also, CONFIG_WIN32 is what we use elsewhere for this test.


r~

Re: [PATCH] qga: Use gethostname() instead of g_get_host_name()

[Michal Privoznik]

On 6/19/20 11:54 PM, Richard Henderson wrote:
> On 6/16/20 1:34 AM, Michal Privoznik wrote:
>> +#ifndef G_OS_WIN32
> 
> Nit: positive tests are easier to reason with and extend than negative tests.
> I would reverse these two blocks and use a positive test for windows.
> 
> Also, CONFIG_WIN32 is what we use elsewhere for this test.
> 
> 
> r~
> 

Fair enough. Do you want me to send v2 or is it something that committer 
can fix before merging?

Michal

Re: [PATCH] qga: Use gethostname() instead of g_get_host_name()

[Philippe Mathieu-Daudé]

Hi Michal,

On 6/16/20 10:34 AM, Michal Privoznik wrote:
> Problem with g_get_host_name() is that on the first call it saves
> the hostname into a global variable and from then on, every
> subsequent call returns the saved hostname. Even if the hostname
> changes. This doesn't play nicely with guest agent, because if
> the hostname is acquired before the guest is set up (e.g. on the
> first boot, or before DHCP) we will report old, invalid hostname.
> 
> Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1845127
> 
> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
> ---
>  qga/commands.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 52 insertions(+), 4 deletions(-)
> 
> diff --git a/qga/commands.c b/qga/commands.c
> index efc8b90281..ce3c2041a6 100644
> --- a/qga/commands.c
> +++ b/qga/commands.c
> @@ -512,14 +512,62 @@ int ga_parse_whence(GuestFileWhence *whence, Error **errp)
>      return -1;
>  }
>  
> +#ifndef HOST_NAME_MAX
> +# ifdef _POSIX_HOST_NAME_MAX
> +#  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
> +# else
> +#  define HOST_NAME_MAX 255
> +# endif
> +#endif
> +
>  GuestHostName *qmp_guest_get_host_name(Error **errp)
>  {
>      GuestHostName *result = NULL;
> -    gchar const *hostname = g_get_host_name();
> -    if (hostname != NULL) {
> -        result = g_new0(GuestHostName, 1);
> -        result->host_name = g_strdup(hostname);
> +    g_autofree char *hostname = NULL;
> +
> +    /*
> +     * We want to avoid using g_get_host_name() because that
> +     * caches the result and we wouldn't reflect changes in the
> +     * host name.
> +     */

I see there is only one g_get_host_name() call in the
codebase, but can we have a generic qemu_get_host_name()
helper implemented in util/oslib-*c instead?

> +
> +#ifndef G_OS_WIN32
> +    long len = -1;
> +
> +#ifdef _SC_HOST_NAME_MAX
> +    len = sysconf(_SC_HOST_NAME_MAX);
> +#endif /* _SC_HOST_NAME_MAX */
> +
> +    if (len < 0) {
> +        len = HOST_NAME_MAX;
>      }
> +
> +    hostname = g_malloc0(len + 1);
> +
> +    if (gethostname(hostname, len) < 0) {
> +        return NULL;
> +    }
> +
> +#else /* G_OS_WIN32 */
> +
> +    wchar_t tmp[MAX_COMPUTERNAME_LENGTH + 1];
> +    DWORD size = G_N_ELEMENTS(tmp);
> +
> +    if (GetComputerNameW(tmp, &size) != 0) {
> +        /*
> +         * Indeed, on Windows retval of zero means failure
> +         * and nonzero means success.
> +         */
> +        hostname = g_utf16_to_utf8(tmp, size, NULL, NULL, NULL);
> +    }
> +#endif /* G_OS_WIN32 */
> +
> +    if (!hostname) {
> +        hostname = g_strdup("localhost");
> +    }
> +
> +    result = g_new0(GuestHostName, 1);
> +    result->host_name = g_steal_pointer(&hostname);
>      return result;
>  }
>  
> 

Re: [PATCH] qga: Use gethostname() instead of g_get_host_name()

[Michal Privoznik]

On 6/22/20 12:14 PM, Philippe Mathieu-Daudé wrote:
> Hi Michal,
> 
> On 6/16/20 10:34 AM, Michal Privoznik wrote:
>> Problem with g_get_host_name() is that on the first call it saves
>> the hostname into a global variable and from then on, every
>> subsequent call returns the saved hostname. Even if the hostname
>> changes. This doesn't play nicely with guest agent, because if
>> the hostname is acquired before the guest is set up (e.g. on the
>> first boot, or before DHCP) we will report old, invalid hostname.
>>
>> Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1845127
>>
>> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
>> ---
>>   qga/commands.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++----
>>   1 file changed, 52 insertions(+), 4 deletions(-)
>>
>> diff --git a/qga/commands.c b/qga/commands.c
>> index efc8b90281..ce3c2041a6 100644
>> --- a/qga/commands.c
>> +++ b/qga/commands.c
>> @@ -512,14 +512,62 @@ int ga_parse_whence(GuestFileWhence *whence, Error **errp)
>>       return -1;
>>   }
>>   
>> +#ifndef HOST_NAME_MAX
>> +# ifdef _POSIX_HOST_NAME_MAX
>> +#  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
>> +# else
>> +#  define HOST_NAME_MAX 255
>> +# endif
>> +#endif
>> +
>>   GuestHostName *qmp_guest_get_host_name(Error **errp)
>>   {
>>       GuestHostName *result = NULL;
>> -    gchar const *hostname = g_get_host_name();
>> -    if (hostname != NULL) {
>> -        result = g_new0(GuestHostName, 1);
>> -        result->host_name = g_strdup(hostname);
>> +    g_autofree char *hostname = NULL;
>> +
>> +    /*
>> +     * We want to avoid using g_get_host_name() because that
>> +     * caches the result and we wouldn't reflect changes in the
>> +     * host name.
>> +     */
> 
> I see there is only one g_get_host_name() call in the
> codebase, but can we have a generic qemu_get_host_name()
> helper implemented in util/oslib-*c instead?

Sure. Let me post a v2 so that I can include Richard's suggestion too.

Michal