Intel 537 modem, not working after new kernel update (2.6.10-1.770_FC3)

suro · 04-20-2005, 11:46 AM

After updating kernel version 2.6.10-1.770_FC3 my modem stopped working, I tried to install the driver, it seemed to be installed (when typing make 537 it showed some errors, see below) but when I open kppp and dial it hangs on Initializing Modem, on the log window "Expecting: OK".
/dev/modem exists and has a link to /dev/537 everything seems to be fine but does not work, before updating kernel the modem was woking fine.

Here is the compilation process.

Code:

[root@localhost intel537]# make 537
   Module precompile check
   Current running kernel is: 2.6.10-1.770_FC3
   /lib/modules...   autoconf.h exists
diff: /boot/vmlinuz.autoconf.h: No such file or directory
   autoconf.h matches running kernel
diff: /boot/vmlinuz.version.h: No such file or directory
   version.h matches running kernel
2.6.10-1.770_FC3
make[1]: Entering directory `/usr/local/src/intel537/coredrv'
make -C /lib/modules/2.6.10-1.770_FC3/build SUBDIRS=/usr/local/src/intel537/coredrv modules
make[2]: Entering directory `/lib/modules/2.6.10-1.770_FC3/build'
  CC [M]  /usr/local/src/intel537/coredrv/coredrv.o
/usr/local/src/intel537/coredrv/coredrv.c: In function `softcore_init_struct':
/usr/local/src/intel537/coredrv/coredrv.c:336: warning: assignment from incompatible pointer type
/usr/local/src/intel537/coredrv/coredrv.c: In function `hamproc_write':
/usr/local/src/intel537/coredrv/coredrv.c:660: warning: ignoring return value of `copy_from_user', declared with attribute warn_unused_result
/usr/local/src/intel537/coredrv/coredrv.c: At top level:
/usr/local/src/intel537/coredrv/coredrv.c:754: warning: initialization from incompatible pointer type
/usr/local/src/intel537/coredrv/coredrv.c:755: warning: initialization from incompatible pointer type
/usr/local/src/intel537/coredrv/coredrv.c: In function `dspdrv_CommRamISR':
/usr/local/src/intel537/coredrv/coredrv.c:877: warning: function declaration isn't a prototype
  CC [M]  /usr/local/src/intel537/coredrv/clmmain.o
  CC [M]  /usr/local/src/intel537/coredrv/rts.o
/usr/local/src/intel537/coredrv/rts.c: In function `RTS_Enable':
/usr/local/src/intel537/coredrv/rts.c:118: warning: assignment from incompatible pointer type
  CC [M]  /usr/local/src/intel537/coredrv/task.o
/usr/local/src/intel537/coredrv/task.c: In function `RSATaskYield':
/usr/local/src/intel537/coredrv/task.c:144: warning: `interruptible_sleep_on' is deprecated (declared at include/linux/wait.h:310)
  CC [M]  /usr/local/src/intel537/coredrv/uart.o
/usr/local/src/intel537/coredrv/uart.c: In function `UART_msr_wait':
/usr/local/src/intel537/coredrv/uart.c:94: warning: `interruptible_sleep_on' is deprecated (declared at include/linux/wait.h:310)
  CC [M]  /usr/local/src/intel537/coredrv/wwh_dflt.o
  CC [M]  /usr/local/src/intel537/coredrv/locks.o
  CC [M]  /usr/local/src/intel537/coredrv/softserial_io.o
/usr/local/src/intel537/coredrv/softserial_io.c: In function `softserial_write':
/usr/local/src/intel537/coredrv/softserial_io.c:94: warning: ignoring return value of `copy_from_user', declared with attribute warn_unused_result
  CC [M]  /usr/local/src/intel537/coredrv/softserial_ioctl.o
  CC [M]  /usr/local/src/intel537/coredrv/softserial.o
/usr/local/src/intel537/coredrv/softserial.c: In function `softserial_register_tty':
/usr/local/src/intel537/coredrv/softserial.c:125: warning: assignment from incompatible pointer type
  CC [M]  /usr/local/src/intel537/coredrv/afedsp_int.o
/usr/local/src/intel537/coredrv/afedsp_int.c:48: warning: function declaration isn't a prototype
/usr/local/src/intel537/coredrv/afedsp_int.c:61: warning: initialization from incompatible pointer type
/usr/local/src/intel537/coredrv/afedsp_int.c:65: warning: function declaration isn't a prototype
/usr/local/src/intel537/coredrv/afedsp_int.c:113: warning: conflicting types for built-in function 'malloc'
/usr/local/src/intel537/coredrv/afedsp_int.c: In function `afe_Write':
/usr/local/src/intel537/coredrv/afedsp_int.c:417: warning: ignoring return value of `copy_from_user', declared with attribute warn_unused_result
/usr/local/src/intel537/coredrv/afedsp_int.c: In function `afe_Read':
/usr/local/src/intel537/coredrv/afedsp_int.c:437: warning: ignoring return value of `copy_to_user', declared with attribute warn_unused_result
/usr/local/src/intel537/coredrv/afedsp_int.c: At top level:
/usr/local/src/intel537/coredrv/afedsp_int.c:454: warning: initialization from incompatible pointer type
  LD [M]  /usr/local/src/intel537/coredrv/Intel537.o
  Building modules, stage 2.
  MODPOST
Warning: could not find /usr/local/src/intel537/coredrv/.537core.lib.cmd for /usr/local/src/intel537/coredrv/537core.lib
  CC      /usr/local/src/intel537/coredrv/Intel537.mod.o
  LD [M]  /usr/local/src/intel537/coredrv/Intel537.ko
make[2]: Leaving directory `/lib/modules/2.6.10-1.770_FC3/build'
make[1]: Leaving directory `/usr/local/src/intel537/coredrv'
[root@localhost intel537]# make install
rm -f /etc/hamregistry.bin
bash 537_inst
running kernel 2.6.10-1.770_FC3
installing hamregistry, used for persistant storage
installing usrsound, a soft buzzer
installing 537 module
redhat 537_boot rc2.d, rc3.d and rc5.d scripts
starting module and utilities
537_inst: line 395: depmod: command not found
error loading Intel537
537_boot: line 55: rmmod: command not found
done
[root@localhost intel537]# su -
[root@localhost ~]# cd /usr/local/src/intel537/
[root@localhost intel537]# make install
rm -f /etc/hamregistry.bin
bash 537_inst
running kernel 2.6.10-1.770_FC3
installing hamregistry, used for persistant storage
installing usrsound, a soft buzzer
installing 537 module
redhat 537_boot rc2.d, rc3.d and rc5.d scripts
starting module and utilities
done

Any help is appreciated :-)

P.S. I have tried both
intel-537ep-2.60.80.0-fc2-up.tgz and
intel-537ep-2.60.80.0-fc2-smp.tgz
drivers

osvaldomarques · 04-21-2005, 02:25 AM

Hi Suro,

I found you question yesterday, when I was also looking for an answer. After googling for almost 2 days, I realized that problem occurred between the kernel versions 2.6.9 and 2.6.10. After examine the differences between the serial drivers of both versions, I made a patch to the ep537 and it now works.

The patch is the following:

Code:

diff -ur ../intel-537EP_secure-2.60.80.0/coredrv/softserial.c ./coredrv/softserial.c
--- ../intel-537EP_secure-2.60.80.0/coredrv/softserial.c        2004-07-30 17:57:27.000000000 -0300
+++ ./coredrv/softserial.c      2005-04-21 01:43:34.638550808 -0300
@@ -47,6 +47,14 @@

 //=============================================================================
 extern void core_init_module(void);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+int  softserial_write_wrapper  (struct tty_struct *stt, const unsigned char *buf, int count)
+{
+       return(softserial_write(stt, 0, buf, count));
+}
+#endif
+
 int init_module(void)
 {
    int rc = 0;
@@ -122,7 +130,11 @@

    G.softserial_tty_driver.open            = softserial_open;
    G.softserial_tty_driver.close           = softserial_close;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+   G.softserial_tty_driver.write           = softserial_write_wrapper;
+#else
    G.softserial_tty_driver.write           = softserial_write;
+#endif
    G.softserial_tty_driver.put_char        = softserial_put_char;
    G.softserial_tty_driver.flush_chars     = softserial_flush_chars;
    G.softserial_tty_driver.write_room      = softserial_write_room;

I recommend you the following steps:
- download from intel the uncompiled driver ("Modem Chipset Driver (Uncompiled) - INTEL-537EP-2.60.80.0.TGZ").
- uncompress and untar it on any directory of your preference;
- save the above text as "patch-intel-537ep-kernel2610.diff" on any directory of your preference;
- cd to the directory created when you untarred the driver;
- execute the following command:

Code:

patch -p0 < "directory-of-the-patch/"patch-intel-537ep-kernel2610.diff

After patching, follow the instructions of the "readme.txt"

Speaking deeply, the interface to write to serial devices was changed in kernel 2.6.10. It was eliminated one parameter of the call for a function in the proprietary part of the driver. Then, the data to be sent to the modem is out of position and the modem driver receives just garbage. This patch creates a wrapper to the function, enabling to receive 3 parameters and send the 4 expected by the driver modem.

Good luck!

Osvaldo.

suro · 04-21-2005, 08:52 AM

Hey man thanks a lot
I'll try that now and let you know :-)

Edit: It works, but now the system unexpectedly hangs, I used to have that sort of things but It used to hang when I was dialing or disconnecting, not when I was already connected and surfing...

Anyway it's better than without a modem :-)

osvaldomarques · 04-21-2005, 04:39 PM

Hi Suro,

Yesterday, when testing the theory that the problem was with 2.6.10, booting with a 2.6.5 which I had on hand, after five minutes surfing, the system hung. If that was not happened, I would released the system with the old kernel.

So, we need to discover if there is some previous version which works and, if needed create another patch to fix the 2.6.10 issues.

I'll keep you informed.

Osvaldo

osvaldomarques · 04-21-2005, 06:15 PM

Hi Suro,

It's me again! I came back to Intel and found a revision for the driver which I didn't see yesterday. It's the uncompiled 2.60.80.1. There are several changes on the driver which may affect the problem of system hang up. The changes invalidate my previous patch, so I wrote another. Today I have not an available machine to test it, so could you try it?
There it goes:

Code:

diff -ur ../intel-537EP_secure-2.60.80.1-sv/coredrv/softserial.c ./coredrv/softserial.c
--- ../intel-537EP_secure-2.60.80.1-sv/coredrv/softserial.c     2004-10-15 12:49:27.000000000 -0300
+++ ./coredrv/softserial.c      2005-04-21 19:42:30.000000000 -0300
@@ -47,6 +47,14 @@
 extern void core_cleanup_module(void);
 //=============================================================================
 extern void core_init_module(void);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+int softserial_write_wrapper(struct tty_struct *stt, const unsigned char *buf, int count)
+{
+       return(softserial_write(stt, 0, buf, count));
+}
+#endif
+
 static int __init init_serial(void)
 {
    int rc = 0;
@@ -138,7 +146,11 @@

    G.softserial_tty_driver->open            = softserial_open;
    G.softserial_tty_driver->close           = softserial_close;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+   G.softserial_tty_driver->write           = softserial_write_wrapper;
+#else
    G.softserial_tty_driver->write           = softserial_write;
+#endif
    G.softserial_tty_driver->put_char        = softserial_put_char;
    G.softserial_tty_driver->flush_chars     = softserial_flush_chars;
    G.softserial_tty_driver->write_room      = softserial_write_room;

Good luck!

Osvaldo.

suro · 04-22-2005, 10:18 AM

Hi Osvaldo
I've downloaded the 2.60.80.1 driver installed your new script but the result is the same, the system has hung after about 2 minutes connected to the internet, so there's no luck.

osvaldomarques · 04-23-2005, 07:06 PM

Hi Suro,

Today I gave access to the machine with this modem again. I visited all the warnings from the compilation with intention of eliminate it with. I work with "c" for more than 10 years and I know that if the c compiler tell you a warning, it is better to look at. My policy is to eliminate even the fool warning to maintain the compilation clean. If we have 100 warnings and after some editing we get one more, a serious one, we wouldn't see it. And, if we don't pay attention to the warnings, we can be in trouble as the first patch I wrote was indicated by one warning.

So, after this blah-blah-blah, I started to rewrite some portions of the sources to eliminate the warnings, one by one, until the last one, in the line 454 of afedsp_int.c. It's a warning like the first. Worst, it can be the responsible for the system hang-ups. This function might be useful in the kernel 2.4 series but it's dummy on 2.6. Instead of remove the warning, I disarmed the bomb inside the function. This machine is already connected to internet for 2:25h at 49333 bps. I'm downloading the kernel patch 2.4.30 for testing the connection. The patch is as follows:

Code:

diff -ur ../intel-537EP_secure-2.60.80.1/coredrv/afedsp_int.c ./coredrv/afedsp_int.c
--- ../intel-537EP_secure-2.60.80.1/coredrv/afedsp_int.c        2004-10-15 11:00:43.000000000 -0300
+++ ./coredrv/afedsp_int.c      2005-04-23 18:30:18.000000000 -0300
@@ -35,7 +35,7 @@
 #include "lock_lin.h"

 static void dsp_function(unsigned long x);
-static void sound_function(unsigned long x);
+static void sound_function(void *x);
 DECLARE_TASKLET(dsp_tasklet, dsp_function, 0);
 DECLARE_WAIT_QUEUE_HEAD(sound_wait);
 struct file_operations proc_afe_operations;
@@ -44,7 +44,7 @@


 void RTS_Task(void *ptr);
-asmlinkage void linux_schedule()
+asmlinkage void linux_schedule(void)
 {
 schedule();
 }
@@ -61,7 +61,7 @@
 DECLARE_WORK(sound_wq, sound_function, 0);
 #endif

-asmlinkage void schedule_precdialer()
+asmlinkage void schedule_precdialer(void)
 {
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,0)
       schedule_task(&precdialer_task);
@@ -339,7 +339,7 @@
     MT_ReleaseLock_mt_dsp();
 }

-static void sound_function(unsigned long x)
+static void sound_function(void *x)
 {
    SoundTask((void*)&devAfe->SysAudio);
 }
@@ -410,12 +410,12 @@
 static ssize_t afe_Write(struct file * file, char * buf,
                        size_t count, loff_t *ppos)
 {
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,5,0)
     char tmpbuf[512];
     if (file->f_flags & O_NONBLOCK) {
         return -EAGAIN;
     }
     copy_from_user(tmpbuf, buf, count);
-#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,5,0)
     circular_buffer_put(write_buffer, tmpbuf, count);
 #endif
     return count;
diff -ur ../intel-537EP_secure-2.60.80.1/coredrv/coredrv.c ./coredrv/coredrv.c
--- ../intel-537EP_secure-2.60.80.1/coredrv/coredrv.c   2004-10-15 05:14:12.000000000 -0300
+++ ./coredrv/coredrv.c 2005-04-23 18:14:25.000000000 -0300
@@ -737,14 +737,14 @@

 static int writeQSem = 0, readQSem = 0;
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
-void WakeTask(void)
+void WakeTask(void *x)
 {
    wake_up_interruptible(&persistReadQ);
    readQSem = 0;
 }
 static int *pkt_to_app1;
 static int timeout1;
-void WakeTask2(void)
+void WakeTask2(void *x)
 {
    while((*pkt_to_app1 == TRUE) && timeout1 > 0) //while pkt hasnt been read yet.
    {
@@ -874,7 +874,7 @@

 asmlinkage void dspdrv_CommRamISR(void)
 {
-    extern asmlinkage void kScheduleDPC();
+    extern asmlinkage void kScheduleDPC(void);
     // clear DSP interrupt
     #if !defined (TARGET_SELAH)
     dspdrv_clear_dsp_interrupt();
diff -ur ../intel-537EP_secure-2.60.80.1/coredrv/softserial.c ./coredrv/softserial.c
--- ../intel-537EP_secure-2.60.80.1/coredrv/softserial.c        2004-10-15 12:49:27.000000000 -0300
+++ ./coredrv/softserial.c      2005-04-23 17:57:46.000000000 -0300
@@ -47,6 +47,14 @@
 extern void core_cleanup_module(void);
 //=============================================================================
 extern void core_init_module(void);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+int softserial_write_wrapper(struct tty_struct *stt, const unsigned char *buf, int count)
+{
+       return(softserial_write(stt, 0, buf, count));
+}
+#endif
+
 static int __init init_serial(void)
 {
    int rc = 0;
@@ -138,7 +146,11 @@

    G.softserial_tty_driver->open            = softserial_open;
    G.softserial_tty_driver->close           = softserial_close;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,10)
+   G.softserial_tty_driver->write           = softserial_write_wrapper;
+#else
    G.softserial_tty_driver->write           = softserial_write;
+#endif
    G.softserial_tty_driver->put_char        = softserial_put_char;
    G.softserial_tty_driver->flush_chars     = softserial_flush_chars;
    G.softserial_tty_driver->write_room      = softserial_write_room;

I hope you enjoy it.

Osvaldo.
P.S. I saw today your new thread looking for a new softmodem. I know it's a nightmare!

suro · 04-24-2005, 03:12 AM

Hi Osvaldo

I tried to patch your new script and got this.

Code:

[root@localhost intel-537EP_secure-2.60.80.1]# patch -p0 < patch.diff
patching file ./coredrv/afedsp_int.c
Hunk #3 succeeded at 61 with fuzz 2.
Hunk #5 FAILED at 410.
1 out of 5 hunks FAILED -- saving rejects to file ./coredrv/afedsp_int.c.rej
patching file ./coredrv/coredrv.c
Hunk #1 FAILED at 737.
1 out of 2 hunks FAILED -- saving rejects to file ./coredrv/coredrv.c.rej
patching file ./coredrv/softserial.c

So I've changed the files manually where they were rejected, then I removed the whole function

Code:

struct file_operations proc_afe_operations = {
    write:	 afe_Write,
    read:	 afe_Read,
    poll:	 afe_Poll,
    open:	 afe_Open,
    release: afe_Release,
};

from afedsp_int.c, was that what you were saying?

So the modem compiled with some errors, worked, but still the system is hanging.
So, if it's not hard for you could you just send me the sources of the files that you have changed?

Thanks.

osvaldomarques · 04-24-2005, 03:54 AM

Hi Suro,

I uploaded the modified sources and the patch to the following address:
537ep-patch

Unpack a fresh driver then save this tar.gz file in the created directory and uncompress it into the directory.

Good luck!

Osvaldo.

By the way, I disconnected the call after 4 hours without any glitch.

suro · 04-24-2005, 04:13 AM

Hey man, now it works just perfect.
Now I can sit back, relax and continue coding on php while being connected to the internet.
You've helped not only me but many many Intel modem users.

Thanks a lot :-)

suro · 04-24-2005, 06:47 AM

I'm sorry to say this Osvaldo, but after I sent the last reply to you the system hung. At that time I waited for about 30 minutes and It didn't hang and I thought that that was it, the problem was solved, but nope I was wrong

I have noticed that when you're doing multiple queries, like opening multiple web pages at the same time at that time the system hangs, but when downloading only one file, the system may not hang.

So, after those 4 hours of surfing on the internet, have you tried to reconnect and try again, hasn't your system hung after that?

osvaldomarques · 04-24-2005, 11:12 AM

Hi Suro,
Sorry to know about it! During that four hours I tried to surf while downloading but I didn't push too hard. And, after disconnect, I didn't try to reconnect again. Now I'm trying to surf but I don't know about any heavy weight page to load.
I have several pages of New York Times, linmodem.org and I disconnected after 10 minutes and reconnected it again. The system is steady until now, after 13 minutes from reconnection.
Do you have some clue how to provoke the crash? Can you point me to some page you access? I will maintain the system up during the afternoon and evening as, at Sundays I have to pay just one impulse for each telephone call.
Waiting for your input,

Osvaldo.

suro · 04-24-2005, 11:40 AM

Well I would say that it doesn't matter what page I load heavy or not. But the system has hung 2 times when I tried to open www.hotscripts.com website.
And as I think it doesn't matter what page I load, sooner or later it crashes.

P.S. Must I type make clean before make 537?

osvaldomarques · 04-24-2005, 11:47 AM

I can't say it but I always do make clean. I'll try to access hotscripts now. The second connection is already for 52 minutes.

See you soon,

Osvaldo.

suro · 04-24-2005, 01:50 PM

P.S. Exactly which kernel are you running?
Mine is 2.6.10-1.770_FC3 and if yours is different then maybe that's the problem?