nvidia 2.5.40+ patch?

Chris Wedgwood · 2002-10-07 01:48 UTC

Quoted 2 lines · Bob McElrath

I applied these patches (I did not disable preempt, and how do you enable stack frame pointer support?)

You need preempt to get this debugging.

Quoted 1 line

and there are 2 oopses in bootup:

Oopses or warnings? There are plenty of locking violations, you see these during init right now, but there they do little harm. Once the system in up --- then they are a problem.

Quoted 6 lines

    Trace; c01387ca <slabinfo_write+32a/6c0>
    Trace; c0213e05 <blk_cleanup_queue+105/180>
    Trace; c0213e97 <blk_init_queue+17/2c0>
    Trace; c0225fd9 <save_match+c9/130>
    Trace; c022d9d0 <do_ide_request+0/30>
    Trace; c022620e <init_irq+1ce/560>

Quoted 8 lines

    Trace; c022dd80 <ide_intr+0/180>
    Trace; c0226678 <hwif_init+d8/260>
    Trace; c0225ec4 <probe_hwif_init+24/70>
    Trace; c0235530 <ide_setup_pci_device+50/80>
    Trace; c02251f6 <generic_mii_ioctl+1286/1c70>
    Trace; c010507a <stext+7a/1e0>
    Trace; c0105040 <stext+40/1e0>
    Trace; c0105625 <show_regs+165/170>

IDE warning, know problem ... fix is pending

Quoted 8 lines

    Trace; c0252cb2 <usb_hub_port_disable+482/8c0>
    Trace; c0115ac2 <schedule+192/300>
    Trace; c011c8b4 <reparent_to_init+e4/180>
    Trace; c0253035 <usb_hub_port_disable+805/8c0>
    Trace; c0115c66 <preempt_schedule+36/50>
    Trace; c0115c80 <default_wake_function+0/a0>
    Trace; c0253000 <usb_hub_port_disable+7d0/8c0>
    Trace; c0105625 <show_regs+165/170>

USB people know about this and many others, fixes sure to come soon I guess.

Quoted 7 lines

However I cannot get the nvidia driver to compile with Chris' patch:
(0)<mcelrath@navi:/usr/src/NVIDIA_kernel-1.0-3123.cw> sudo insmod ./NVdriver ./NVdriver: unresolved symbol create_workqueue
./NVdriver: unresolved symbol destroy_workqueue
./NVdriver: unresolved symbol flush_workqueue
./NVdriver: unresolved symbol queue_work
./NVdriver:

What kernel? My patch requires stuff only in bk://linux.bkbits.net/linux-2.5 at present, the next release will have the workqueue stuff too.

--cw

Bob McElrath · 2002-10-07 03:13 UTC

Quoted 2 lines · Chris Wedgwood [[email protected]] wrote:

My changes are pretty small and don't oops. Like I said though, something just isn't quite right though :)

Yes indeed, your patch works fine, but as you observed, it hangs sometimes. My guess is the work queue work isn't getting done on time.

If you give the [nvisr/0] process a negative priority, the stalls go away. I just played through a level of Chromium BSU and it performed flawlessly (with the nvisr process at -20 priority). With the latest bk tree there are also no oopses.

I'm not sure how to solve this in a general way. But the nvisr work isn't getting done quickly enough. I think under the task queue system the bottom-half interrupt work got scheduled as immediate (i.e. it's guaranteed to be the next work run). But under work queues it enters the scheduler like any other process, so several timeslices can pass before it gets done.

Really, the nvisr process should be SCHED_FIFO. (i think) Or whatever it takes to ensure that when it is given work, it gets scheduled immediately.

Correct me if I'm wrong...I know diddly squat about task/work queues/bh isr's...

I'm re-sending your patch against the nvidia driver with this post and copying it to lkml, since it works great and others might want to see what the code does.

Cheers,
-- Bob

Bob McElrath
Univ. of Wisconsin at Madison, Department of Physics

"The purpose of separation of church and state is to keep forever from these shores the ceaseless strife that has soaked the soil of Europe in blood for centuries." -- James Madison

Index: Makefile
===================================================================
RCS file: /var/cvs/nvkern/Makefile,v
retrieving revision 1.1.1.1
retrieving revision 1.2

Makefile +2 -2

diff -u -r1.1.1.1 -r1.2
@@ -57,9 +57,9 @@
 
 # allow specification of alternate include file tree on command line and extra defines
 ifdef SYSINCLUDE
-INCLUDES += -I$(SYSINCLUDE)
+INCLUDES += -I$(SYSINCLUDE) -I$(KERNINC)/../arch/i386/mach-generic
 else
-INCLUDES += -I$(KERNINC)
+INCLUDES += -I$(KERNINC) -I$(KERNINC)/../arch/i386/mach-generic
 endif
 
 DEFINES+=$(EXTRA_DEFINES)

Index: nv-linux.h
===================================================================
RCS file: /var/cvs/nvkern/nv-linux.h,v
retrieving revision 1.1.1.1
retrieving revision 1.2

nv-linux.h +2 -7

diff -u -r1.1.1.1 -r1.2
@@ -36,11 +36,6 @@
 #  error This driver does not support 2.3.x development kernels!
 #elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 5, 0)
 #  define KERNEL_2_4
-#elif LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0)
-#  error This driver does not support 2.5.x development kernels!
-#  define KERNEL_2_5
-#else
-#  error This driver does not support 2.6.x or newer kernels!
 #endif
 
 #if defined (CONFIG_SMP) && !defined (__SMP__)

@@ -51,7 +46,7 @@
 #include <linux/errno.h>            /* error codes                      */
 #include <linux/stddef.h>           /* NULL, offsetof                   */
 #include <linux/wait.h>             /* wait queues                      */
-#include <linux/tqueue.h>           /* struct tq_struct                 */
+#include <linux/workqueue.h>
 
 #include <linux/slab.h>             /* kmalloc, kfree, etc              */
 #include <linux/vmalloc.h>          /* vmalloc, vfree, etc              */

@@ -194,7 +189,7 @@
     nv_alloc_t *alloc_queue;
 
     // bottom half interrupt handler info; per device
-    struct tq_struct *bh;
+    struct work_struct *work;
 
     U032 vblank_notifier;
     U032 waiting_for_vblank;

Index: nv.c
===================================================================
RCS file: /var/cvs/nvkern/nv.c,v
retrieving revision 1.1.1.1
retrieving revision 1.2

nv.c +36 ~~-32~~

184 lines not on this page · raw message

Index: nv.h
===================================================================
RCS file: /var/cvs/nvkern/nv.h,v
retrieving revision 1.1.1.1
retrieving revision 1.2

nv.h +5 -0

diff -u -r1.1.1.1 -r1.2
@@ -195,6 +195,11 @@
     U032 agp_buffers;
     U032 agp_teardown;
 
+    /*
+     * CW:XXX.  I'm not sure *how* the value below ever gets
+     * decremented, perhaps inside the magic binary logic in which case
+     * we need to keep it about
+     */
     /* keep track of any pending bottom-halves */
     int bh_count;

Index: os-interface.c
===================================================================
RCS file: /var/cvs/nvkern/os-interface.c,v
retrieving revision 1.1.1.1
retrieving revision 1.2

os-interface.c +8 -4

diff -u -r1.1.1.1 -r1.2
@@ -27,7 +27,7 @@
 
 BOOL os_is_administrator(PHWINFO pDev)
 {
-    return suser();
+    return (capable(CAP_SYS_ADMIN) != 0);
 }
 
 U032 os_get_page_size(VOID)

@@ -1141,7 +1141,7 @@
     uaddr = *priv;
 
     /* finally, let's do it! */
-    err = remap_page_range( (size_t) uaddr, (size_t) paddr, size_bytes, 
+    err = remap_page_range( kaddr, (size_t) uaddr, (size_t) paddr, size_bytes,
                             PAGE_SHARED);
 
     if (err != 0)

@@ -1177,7 +1177,7 @@
     uaddr = *priv;
 
     /* finally, let's do it! */
-    err = remap_page_range( (size_t) uaddr, (size_t) start, size_bytes, 
+    err = remap_page_range( *priv, (size_t) uaddr, (size_t) start, size_bytes, 
                             PAGE_SHARED);
 
     if (err != 0)

@@ -1560,8 +1560,10 @@
 
     *pPriv_data = data;
     
+#if 0 /* CW:XXX silence */
     printk(KERN_INFO "NVRM: AGPGART: allocated %lu pages\n",
            (unsigned long) PageCount);
+#endif
     return RM_OK;
 
 fail:

@@ -1593,7 +1595,7 @@
 
     agp_addr = agpinfo.aper_base + (agp_data->offset << PAGE_SHIFT);
 
-    err = remap_page_range(vma->vm_start, (size_t) agp_addr, 
+    err = remap_page_range(vma, vma->vm_start, (size_t) agp_addr, 
                            agp_data->num_pages << PAGE_SHIFT,
 #if defined(NVCPU_IA64)
                            vma->vm_page_prot);

@@ -1686,7 +1688,9 @@
         (*(agp_ops.unbind_memory))(ptr);
         (*(agp_ops.free_memory))(ptr);
 
+#if 0 /* CW:XXX silence */
         printk(KERN_INFO "NVRM: AGPGART: freed %ld pages\n", (unsigned long)pages);
+#endif
         PRINT_AGPGART_TABLE();
         return RM_OK;
     }

Bob McElrath · 2002-10-07 05:33 UTC

Quoted 4 lines · Bob McElrath [[email protected]] wrote:

If you give the [nvisr/0] process a negative priority, the stalls go away. I just played through a level of Chromium BSU and it performed flawlessly (with the nvisr process at -20 priority). With the latest bk tree there are also no oopses.

So it seems the way to do this is:

    my_workqueue = create_workqueue("mywork");
    if (!my_workqueue) {
        printk("Unable to create workqueue\n");
        return -EIO;
    }
    pri.sched_priority = 1; /* BM: any nonzero should do, is there any reason 
                               to chose a larger priority? */
    rc = security_ops->task_setscheduler(isr_workqueue->cpu_wq[0]->pid, SCHED_FIFO, &pri);
    if(rc < 0) {
        printk("Unable to set real-time priority for NV (err=%d)\n", rc);
    }

The above code doesn't work though because the struct workqueue_struct is hidden away in workqueue.c, invisible to modules. The line:

    isr_workqueue->cpu_wq[0]->pid

bombs because the struct is not defined (dereferencing pointer to incomplete type). Is there a better way to get the pid of a workqueue I just created? Or for that matter is there a better way to set the scheduler for a workqueue I just created? I imagine many drivers would want to do this...

I notice the sched_setscheduler system call is not available in any kernel header. What is the proper procedure for calling such a function from a module?

Cheers,
-- Bob

Bob McElrath
Univ. of Wisconsin at Madison, Department of Physics

"The purpose of separation of church and state is to keep forever from these shores the ceaseless strife that has soaked the soil of Europe in blood for centuries." -- James Madison