Fw: 2.5.75 - ppa unload PATCH and remaining problem

Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Andrew Morton]

[-- Attachment #1: Type: text/plain, Size: 4738 bytes --]


Could someone (Mike?) please take a look at the ppa driver,
see what it's doing wrong?

Thanks.


Begin forwarded message:

Date: Sun, 13 Jul 2003 18:47:39 +0400
From: Andrey Borzenkov <arvidjaar@mail.ru>
To: linux-kernel@vger.kernel.org
Subject: 2.5.75 - ppa unload PATCH and remaining problem


rmmod ppa does not remove any attached device nodes and gives:

ppa: Version 2.07 (for Linux 2.4.x)
ppa: Found device at ID 4, Attempting to use EPP 32 bit
ppa: Found device at ID 4, Attempting to use PS/2
ppa: Communication established with ID 4 using PS/2
scsi0 : Iomega VPI0 (ppa) interface
  Vendor: iomega    Model: jaz 2GB           Rev: E.17
  Type:   Direct-Access                      ANSI SCSI revision: 02
Attached scsi removable disk sda at scsi0, channel 0, id 4, lun 0
Attached scsi generic sg0 at scsi0, channel 0, id 4, lun 0,  type 0
sda: Spinning up disk....ready
SCSI device sda: 2091050 512-byte hdwr sectors (1071 MB)
sda: Write Protect is off
sda: Mode Sense: 39 00 10 08
SCSI device sda: drive cache: write through
SCSI device sda: 2091050 512-byte hdwr sectors (1071 MB)
sda: Write Protect is off
sda: Mode Sense: 39 00 10 08
SCSI device sda: drive cache: write through
 /dev/scsi/host0/bus0/target4/lun0: p4
Releasing ppa0
Iomega VPI0 (ppa) interface did not call scsi_unregister
Call Trace:
 [<d2d2fcf6>] exit_this_scsi_driver+0xb6/0xfa [ppa]
 [<c013f0ad>] sys_delete_module+0x11d/0x150
 [<c015787b>] do_munmap+0x13b/0x1a0
 [<c0157923>] sys_munmap+0x43/0x70
 [<c011d920>] do_page_fault+0x0/0x4b8
 [<c010b527>] syscall_call+0x7/0xb


adding scsi_unregister to ppa_release does remove all devices and sysfs node 
for ppa but leaves /dev/scsi/host0/bus0/target4 in place and does not remove 
/proc entries; loading ppa again creates second host with /proc/scsi looking 
funny:

{pts/3}% LC_ALL=C ls -l /proc/scsi
total 0
-r--r--r--    1 root     root            0 Jul 13 18:42 device_info
dr-xr-xr-x    2 root     root            0 Jul 13 18:42 ppa/
dr-xr-xr-x    2 root     root            0 Jul 13 18:42 ppa/
-r--r--r--    1 root     root            0 Jul 13 18:42 scsi
dr-xr-xr-x    2 root     root            0 Jul 13 18:42 sg/

attempt to access /proc/scsi after ppa has been unloaded (with my patch or 
without) even after the first time gives:

pts/3}% LC_ALL=C ls -lR /proc/scsi
zsh: segmentation fault  LC_ALL=C ls -F --color=auto -lR /proc/scsi

dmesg ->

Unable to handle kernel paging request at virtual address d2d42320
 printing eip:
c0199360
*pde = 09fa3067
*pte = 00000000
Oops: 0000 [#1]
CPU:    0
EIP:    0060:[<c0199360>]    Tainted: P
EFLAGS: 00010202
EIP is at proc_get_inode+0xb0/0x1a0
eax: 00000000   ebx: c5a3e014   ecx: 00000002   edx: d2d42320
esi: c2f4a4fc   edi: 00000001   ebp: c77afe30   esp: c77afe24
ds: 007b   es: 007b   ss: 0068
Process ls (pid: 2893, threadinfo=c77ae000 task=c76c9000)
Stack: 00000003 c5ab708b c2f4a54b c77afe5c c019cb29 cff93004 000011e7 c2f4a4fc
       c2f4a4fc ffffffea 00000000 fffffff4 cea33084 cea33014 c77afe80 c0173a25
       cea33014 c5ab7004 c77aff30 c5ab7004 c77afef4 c77aff30 cffb08e4 c77afea0
Call Trace:
 [<c019cb29>] proc_lookup+0x169/0x190
 [<c0173a25>] real_lookup+0xb5/0xe0
 [<c0173ebd>] do_lookup+0x6d/0x80
 [<c01744a7>] link_path_walk+0x5d7/0xb30
 [<c014c129>] kmem_cache_alloc+0x99/0x1a0
 [<c0174f92>] __user_walk+0x32/0x50
 [<c016f127>] vfs_lstat+0x17/0x50
 [<c016f724>] sys_lstat64+0x14/0x30
 [<c01796f4>] sys_getdents64+0x74/0xb2
 [<c017972d>] sys_getdents64+0xad/0xb2
 [<c010b527>] syscall_call+0x7/0xb

Code: 83 3a 02 0f 84 aa 00 00 00 c1 e0 05 8d 04 10 f0 ff 80 a0 00
 <6>note: ls[2893] exited with preempt_count 2
Debug: sleeping function called from illegal context at 
include/asm/semaphore.h:119
Call Trace:
 [<c0122bb8>] __might_sleep+0x58/0x70
 [<c01560bb>] remove_shared_vm_struct+0x2b/0x90
 [<c0157cfa>] exit_mmap+0x19a/0x250
 [<c012375c>] mmput+0x8c/0x100
 [<c012868c>] do_exit+0x19c/0x690
 [<c010cbf7>] die+0x137/0x140
 [<c011da6a>] do_page_fault+0x14a/0x4b8
 [<c014c1bb>] kmem_cache_alloc+0x12b/0x1a0
 [<c0199107>] proc_alloc_inode+0x17/0x70
 [<c0199136>] proc_alloc_inode+0x46/0x70
 [<c011d920>] do_page_fault+0x0/0x4b8
 [<c010c54d>] error_code+0x2d/0x40
 [<c018007b>] d_delete+0x15b/0x270
 [<c0199360>] proc_get_inode+0xb0/0x1a0
 [<c019cb29>] proc_lookup+0x169/0x190
 [<c0173a25>] real_lookup+0xb5/0xe0
 [<c0173ebd>] do_lookup+0x6d/0x80
 [<c01744a7>] link_path_walk+0x5d7/0xb30
 [<c014c129>] kmem_cache_alloc+0x99/0x1a0
 [<c0174f92>] __user_walk+0x32/0x50
 [<c016f127>] vfs_lstat+0x17/0x50
 [<c016f724>] sys_lstat64+0x14/0x30
 [<c01796f4>] sys_getdents64+0x74/0xb2
 [<c017972d>] sys_getdents64+0xad/0xb2
 [<c010b527>] syscall_call+0x7/0xb

Any pointers appreciated. It is tainted by nVidia if question arises.

TIA

-andrey


[-- Attachment #2: 2.5.75-ppa-scsi_unregister.patch --]
[-- Type: text/x-diff, Size: 410 bytes --]

--- linux-2.5.75-smp/drivers/scsi/ppa.c.scsi_unregister	2003-06-26 21:41:23.000000000 +0400
+++ linux-2.5.75-smp/drivers/scsi/ppa.c	2003-07-13 18:05:07.000000000 +0400
@@ -76,6 +76,7 @@ int ppa_release(struct Scsi_Host *host)
     int host_no = host->unique_id;
 
     printk("Releasing ppa%i\n", host_no);
+    scsi_unregister(host);
     parport_unregister_device(ppa_hosts[host_no].dev);
     return 0;
 }

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Christoph Hellwig]

On Sun, Jul 13, 2003 at 02:15:56PM -0700, Andrew Morton wrote:
> sda: Mode Sense: 39 00 10 08
> SCSI device sda: drive cache: write through
>  /dev/scsi/host0/bus0/target4/lun0: p4
> Releasing ppa0
> Iomega VPI0 (ppa) interface did not call scsi_unregister

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Andrew Morton]

Christoph Hellwig <hch@infradead.org> wrote:
>
> On Sun, Jul 13, 2003 at 02:15:56PM -0700, Andrew Morton wrote:
> > sda: Mode Sense: 39 00 10 08
> > SCSI device sda: drive cache: write through
> >  /dev/scsi/host0/bus0/target4/lun0: p4
> > Releasing ppa0
> > Iomega VPI0 (ppa) interface did not call scsi_unregister
> 
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Look at Andrey's report.  His patch adds the scsi_unregister() call.

Then the thing oopses.

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Mike Anderson]

I believe this problem is related to /proc/scsi cleanup issues
previously discussed by Alan Stern.

I refreshed Alan's patch and added my host present bit cleanups to the
last series of patches sent to the list.

I do not have a ppa device, but on 2.5.75 I saw a similar error
signature of the proc not getting cleaned up when I use the aic7xxx_old
which uses the legacy registration interfaces also. When I switched to
my patched view the aic7xxx_old cleans up ok. I have other sysfs
cleanups in this patched tree, but I do not believe they are effecting
the legacy proc cleanups.

-andmike
--
Michael Anderson
andmike@us.ibm.com

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Andrew Morton]

Mike Anderson <andmike@us.ibm.com> wrote:
>
> I believe this problem is related to /proc/scsi cleanup issues
> previously discussed by Alan Stern.
> 
> I refreshed Alan's patch and added my host present bit cleanups to the
> last series of patches sent to the list.
> 
> I do not have a ppa device, but on 2.5.75 I saw a similar error
> signature of the proc not getting cleaned up when I use the aic7xxx_old
> which uses the legacy registration interfaces also. When I switched to
> my patched view the aic7xxx_old cleans up ok. I have other sysfs
> cleanups in this patched tree, but I do not believe they are effecting
> the legacy proc cleanups.

OK, thanks.

Where could we get the relevant diffs from for testers?

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Mike Anderson]

I refreshed the patch against 2.6.0-test1 and I am testing it now. I
should have it to Andrey within the hour.

Andrew Morton [akpm@osdl.org] wrote:
> Mike Anderson <andmike@us.ibm.com> wrote:
> >
> > I believe this problem is related to /proc/scsi cleanup issues
> > previously discussed by Alan Stern.
> > 
> > I refreshed Alan's patch and added my host present bit cleanups to the
> > last series of patches sent to the list.
> > 
> > I do not have a ppa device, but on 2.5.75 I saw a similar error
> > signature of the proc not getting cleaned up when I use the aic7xxx_old
> > which uses the legacy registration interfaces also. When I switched to
> > my patched view the aic7xxx_old cleans up ok. I have other sysfs
> > cleanups in this patched tree, but I do not believe they are effecting
> > the legacy proc cleanups.
> 
> OK, thanks.
> 
> Where could we get the relevant diffs from for testers?
> -
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
-andmike
--
Michael Anderson
andmike@us.ibm.com

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Mike Anderson]

Here is the patch rediffed against 2.6.0-test1. I also joined Alan Stern's
patch and my reduction of scsi_host present users as this is the way I
have been testing it.

This patch also applies to scsi-misc-2.5 with 2 offset warnings.

-andmike
--
Michael Anderson
andmike@us.ibm.com


DESC
This is a refresh of a patch by Alan Stern to correct the removal of scsi
procfs host enteries

http://marc.theaimsgroup.com/?l=linux-scsi&m=105545175600506&w=2

This patch also contains a join of another patch that reduces the users
of scsi_host present.

EDESC


 drivers/scsi/NCR53c406a.c    |    5 ++--
 drivers/scsi/aacraid/linit.c |    1 
 drivers/scsi/hosts.c         |    4 +--
 drivers/scsi/scsi_module.c   |    2 -
 drivers/scsi/scsi_priv.h     |    4 +++
 drivers/scsi/scsi_proc.c     |   49 ++++++++++++++++++++++++++++---------------
 6 files changed, 42 insertions(+), 23 deletions(-)

diff -puN drivers/scsi/hosts.c~stern_present drivers/scsi/hosts.c
--- remove-bleed-2.5/drivers/scsi/hosts.c~stern_present	Mon Jul 14 11:02:06 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/hosts.c	Mon Jul 14 11:02:06 2003
@@ -112,7 +112,7 @@ void scsi_free_shost(struct Scsi_Host *s
 		shost->eh_notify = NULL;
 	}
 
-	shost->hostt->present--;
+	scsi_proc_hostdir_rm(shost->hostt);
 	scsi_destroy_command_freelist(shost);
 	kfree(shost);
 }
@@ -219,7 +219,7 @@ struct Scsi_Host *scsi_host_alloc(struct
 	kernel_thread((int (*)(void *))scsi_error_handler, shost, 0);
 	wait_for_completion(&complete);
 	shost->eh_notify = NULL;
-	shost->hostt->present++;
+	scsi_proc_hostdir_add(shost->hostt);
 	return shost;
  fail:
 	kfree(shost);
diff -puN drivers/scsi/scsi_priv.h~stern_present drivers/scsi/scsi_priv.h
--- remove-bleed-2.5/drivers/scsi/scsi_priv.h~stern_present	Mon Jul 14 11:02:06 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/scsi_priv.h	Mon Jul 14 11:02:06 2003
@@ -90,11 +90,15 @@ extern void scsi_exit_queue(void);
 
 /* scsi_proc.c */
 #ifdef CONFIG_PROC_FS
+extern void scsi_proc_hostdir_add(struct scsi_host_template *);
+extern void scsi_proc_hostdir_rm(struct scsi_host_template *);
 extern void scsi_proc_host_add(struct Scsi_Host *);
 extern void scsi_proc_host_rm(struct Scsi_Host *);
 extern int scsi_init_procfs(void);
 extern void scsi_exit_procfs(void);
 #else
+# define scsi_proc_hostdir_add(sht)	do { } while (0)
+# define scsi_proc_hostdir_rm(sht)	do { } while (0)
 # define scsi_proc_host_add(shost)	do { } while (0)
 # define scsi_proc_host_rm(shost)	do { } while (0)
 # define scsi_init_procfs()		(0)
diff -puN drivers/scsi/scsi_proc.c~stern_present drivers/scsi/scsi_proc.c
--- remove-bleed-2.5/drivers/scsi/scsi_proc.c~stern_present	Mon Jul 14 11:02:06 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/scsi_proc.c	Mon Jul 14 11:02:06 2003
@@ -41,6 +41,8 @@
 struct proc_dir_entry *proc_scsi;
 EXPORT_SYMBOL(proc_scsi);
 
+/* Protect sht->present and sht->proc_dir */
+static DECLARE_MUTEX(global_host_template_sem);
 
 static int proc_scsi_read(char *buffer, char **start, off_t offset,
 			  int length, int *eof, void *data)
@@ -77,16 +79,10 @@ out:
 	return ret;
 }
 
-void scsi_proc_host_add(struct Scsi_Host *shost)
+void scsi_proc_hostdir_add(struct scsi_host_template *sht)
 {
-	struct scsi_host_template *sht = shost->hostt;
-	struct proc_dir_entry *p;
-	char name[10];
-
-	if (!sht->proc_info)
-		return;
-
-	if (!sht->proc_dir) {
+	down(&global_host_template_sem);
+	if (!sht->present++) {
 		sht->proc_dir = proc_mkdir(sht->proc_name, proc_scsi);
         	if (!sht->proc_dir) {
 			printk(KERN_ERR "%s: proc_mkdir failed for %s\n",
@@ -95,6 +91,27 @@ void scsi_proc_host_add(struct Scsi_Host
 		}
 		sht->proc_dir->owner = sht->module;
 	}
+	up(&global_host_template_sem);
+}
+
+void scsi_proc_hostdir_rm(struct scsi_host_template *sht)
+{
+	down(&global_host_template_sem);
+	if (!--sht->present && sht->proc_dir) {
+		remove_proc_entry(sht->proc_name, proc_scsi);
+		sht->proc_dir = NULL;
+	}
+	up(&global_host_template_sem);
+}
+
+void scsi_proc_host_add(struct Scsi_Host *shost)
+{
+	struct scsi_host_template *sht = shost->hostt;
+	struct proc_dir_entry *p;
+	char name[10];
+
+	if (!sht->proc_dir)
+		return;
 
 	sprintf(name,"%d", shost->host_no);
 	p = create_proc_read_entry(name, S_IFREG | S_IRUGO | S_IWUSR,
@@ -107,20 +124,18 @@ void scsi_proc_host_add(struct Scsi_Host
 	} 
 
 	p->write_proc = proc_scsi_write_proc;
-	p->owner = shost->hostt->module;
+	p->owner = sht->module;
 }
 
 void scsi_proc_host_rm(struct Scsi_Host *shost)
 {
-	struct scsi_host_template *sht = shost->hostt;
 	char name[10];
 
-	if (sht->proc_info) {
-		sprintf(name,"%d", shost->host_no);
-		remove_proc_entry(name, sht->proc_dir);
-		if (!sht->present)
-			remove_proc_entry(sht->proc_name, proc_scsi);
-	}
+	if (!shost->hostt->proc_dir)
+		return;
+
+	sprintf(name,"%d", shost->host_no);
+	remove_proc_entry(name, shost->hostt->proc_dir);
 }
 
 static int proc_print_scsidevice(struct device *dev, void *data)
diff -puN drivers/scsi/aacraid/linit.c~stern_present drivers/scsi/aacraid/linit.c
--- remove-bleed-2.5/drivers/scsi/aacraid/linit.c~stern_present	Mon Jul 14 12:42:29 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/aacraid/linit.c	Mon Jul 14 12:42:29 2003
@@ -295,7 +295,6 @@ static int aac_detect(Scsi_Host_Template
 			printk(KERN_WARNING "aacraid: unable to register \"aac\" device.\n");
 	}
 
-	template->present = aac_count; /* # of cards of this type found */
 	return aac_count;
 }
 
diff -puN drivers/scsi/NCR53c406a.c~stern_present drivers/scsi/NCR53c406a.c
--- remove-bleed-2.5/drivers/scsi/NCR53c406a.c~stern_present	Mon Jul 14 12:42:29 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/NCR53c406a.c	Mon Jul 14 12:42:29 2003
@@ -450,6 +450,7 @@ static __inline__ int NCR53c406a_pio_wri
 
 static int __init NCR53c406a_detect(Scsi_Host_Template * tpnt)
 {
+	int present = 0;
 	struct Scsi_Host *shpnt = NULL;
 #ifndef PORT_BASE
 	int i;
@@ -522,7 +523,7 @@ static int __init NCR53c406a_detect(Scsi
 
 	DEB(printk("NCR53c406a: using port_base 0x%x\n", port_base));
 
-	tpnt->present = 1;
+	present = 1;
 	tpnt->proc_name = "NCR53c406a";
 
 	shpnt = scsi_register(tpnt, 0);
@@ -576,7 +577,7 @@ static int __init NCR53c406a_detect(Scsi
 	sprintf(info_msg, "NCR53c406a at 0x%x, IRQ %d, %s PIO mode.", port_base, irq_level, fast_pio ? "fast" : "slow");
 #endif
 
-	return (tpnt->present);
+	return (present);
 
 #if USE_DMA
       err_free_irq:
diff -puN drivers/scsi/scsi_module.c~stern_present drivers/scsi/scsi_module.c
--- remove-bleed-2.5/drivers/scsi/scsi_module.c~stern_present	Mon Jul 14 12:42:29 2003
+++ remove-bleed-2.5-andmike/drivers/scsi/scsi_module.c	Mon Jul 14 12:42:29 2003
@@ -33,7 +33,7 @@ static int __init init_this_scsi_driver(
 	INIT_LIST_HEAD(&sht->legacy_hosts);
 
 	sht->detect(sht);
-	if (!sht->present)
+	if (list_empty(&sht->legacy_hosts))
 		return -ENODEV;
 
 	list_for_each_entry(shost, &sht->legacy_hosts, sht_legacy_list) {

_

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Andrew Morton]

Andrey Borzenkov <arvidjaar@mail.ru> wrote:
>
> Well, the reason is simple and is fixed in patch sent two days ago to lkml - 
> fs/partitions/check.c:register_disk does not add .../disc if media capacity 
> == 0 - i.e. no media present.

Please resend a tested, documented patch for that.

Things are slow at present - Linus is away, the big conference is next
week, etc.  So please be patient, keep working this and keep hassling folks
over it, OK?

Re: Fw: 2.5.75 - ppa unload PATCH and remaining problem

[Mike Anderson]

Andrey Borzenkov [arvidjaar@mail.ru] wrote:
> There are still remaining problems. Sorry if they are already known.
> 
> 1. Removing USB memory stick without unmounting file system first results in 
> dead system. It crashes in several seconds with stack (sorry, had to write 
> down manually)
> 
> sync_sb_inodes+0x97/0x2f0
> writeback_inodes
> wb_kupdate
> __pdflush
> schedule_tail
> pdflush_+0x00/0x20
> pdflush_0xf/0x20
> wb_kupdate
> kernel_thread_helper
> 
> Only Alt-Sysrq-B works after that.
> 
> 2.0.0-test1 + this patch + my ppa patch
> 
> This is currently showstopper for me :(
> 

I previously posted a patch to linux-scsi for scsi host and scsi device
ref count cleanups. The thread is available here:
http://marc.theaimsgroup.com/?t=105770318200001&r=1&w=2

I have updated the patch with comments, but still want to finish a change
to scsi_free_sdev and do some testing before I send it out again.

I do not run devfs currently so I have not seen the issue related to
devfs.

-andmike
--
Michael Anderson
andmike@us.ibm.com