[PATCH] Support for 32 bit iwpriv on 64 bits kernels

[PATCH] Support for 32 bit iwpriv on 64 bits kernels

[Dibyajyoti Ghosh]

This patch adds pointer conversion from 32 bit to 64 bit and vice versa,
if the ioctl is issued from 32 bit iwpriv to 64 bit Kernel.

Signed-off-by: Dibyajyoti Ghosh <dibyajyotig@gmail.com>
---
 wireless_tools.29/iwpriv.c | 41 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)

diff --git a/wireless_tools.29/iwpriv.c b/wireless_tools.29/iwpriv.c
index 4172fe2..a0b1b02 100644
--- a/wireless_tools.29/iwpriv.c
+++ b/wireless_tools.29/iwpriv.c
@@ -243,6 +243,7 @@ iw_usage(void)
   fprintf(stderr, "Usage: iwpriv interface [private-command [private-arguments]]\n");
 }
 
+#include<stdio.h>
 /************************* SETTING ROUTINES **************************/
 
 /*------------------------------------------------------------------*/
@@ -259,12 +260,20 @@ set_private_cmd(int		skfd,		/* Socket */
 		int		priv_num)	/* Number of descriptions */
 {
   struct iwreq	wrq;
+  struct compat_iw_point {
+    __u64 pointer;
+    __u16 length;
+    __u16 flags;
+  };
   u_char	buffer[4096];	/* Only that big in v25 and later */
   int		i = 0;		/* Start with first command arg */
   int		k;		/* Index in private description table */
   int		temp;
   int		subcmd = 0;	/* sub-ioctl index */
   int		offset = 0;	/* Space for sub-ioctl index */
+  FILE		*in;
+  char 		buff[4];
+  int		kspace, uspace;
 
   /* Check if we have a token index.
    * Do it now so that sub-ioctl takes precedence, and so that we
@@ -457,6 +466,26 @@ set_private_cmd(int		skfd,		/* Socket */
 	}
     }
 
+  in = popen("getconf LONG_BIT", "r");
+  fgets(buff, sizeof(buff), in);
+  pclose(in);
+  kspace = atoi(buff);
+  uspace = sizeof(char *) * 8;
+
+  struct compat_iw_point *iwp_compat = (struct compat_iw_point *)&wrq.u.data;
+  __u16 len = wrq.u.data.length;
+  __u16 flags = wrq.u.data.flags;
+
+  /* If 32 bit app is running on 64 bit kernel
+   * convert 32 bit iw_point type payload to 64 bit
+   */
+  if(uspace == 32 && kspace == 64)
+    {
+      iwp_compat->pointer = (__u64)((__u32)wrq.u.data.pointer);
+      iwp_compat->length = len;
+      iwp_compat->flags = flags;
+    }
+
   /* Perform the private ioctl */
   if(ioctl(skfd, priv[k].cmd, &wrq) < 0)
     {
@@ -465,6 +494,18 @@ set_private_cmd(int		skfd,		/* Socket */
       return(-1);
     }
 
+  /* If 32 bit app is running on 64 bit kernel, convert the
+   * 64 bit result back to 32 bit to fit the app
+   */
+  if(uspace == 32 && kspace == 64)
+    {
+      flags = iwp_compat->flags;
+      len = iwp_compat->length;
+      wrq.u.data.pointer = (__u32)iwp_compat->pointer;
+      wrq.u.data.length = iwp_compat->length;
+      wrq.u.data.flags = iwp_compat->flags;
+    }
+
   /* If we have to get some data */
   if((priv[k].get_args & IW_PRIV_TYPE_MASK) &&
      (priv[k].get_args & IW_PRIV_SIZE_MASK))
-- 
1.9.1

Re: [PATCH] Support for 32 bit iwpriv on 64 bits kernels

[Johannes Berg]

On Thu, 2016-06-02 at 20:51 +0530, Dibyajyoti Ghosh wrote:
> This patch adds pointer conversion from 32 bit to 64 bit and vice
> versa,
> if the ioctl is issued from 32 bit iwpriv to 64 bit Kernel.
> 
Huh? the kernel should handle that, no? And didn't you just submit a
patch for it, so you break it again with this change?

johannes

Re: [PATCH] Support for 32 bit iwpriv on 64 bits kernels

[Dibyajyoti Ghosh]

Hi

Sorry for the conflict. We misunderstood from our kernel-patch related
conversation that it is you are reluctant to accept that. So we
attempted to accomplish the same through app-side modification. Any
way, sorry again.

However, we are again concentrating on our kernel side modification,
and will update you soon.

Thanks,
Dibyajyoti


On Thu, Jun 2, 2016 at 9:06 PM, Johannes Berg <johannes@sipsolutions.net> wrote:
> On Thu, 2016-06-02 at 20:51 +0530, Dibyajyoti Ghosh wrote:
>> This patch adds pointer conversion from 32 bit to 64 bit and vice
>> versa,
>> if the ioctl is issued from 32 bit iwpriv to 64 bit Kernel.
>>
> Huh? the kernel should handle that, no? And didn't you just submit a
> patch for it, so you break it again with this change?
>
> johannes