: 2.6.24-rc5: mtd: PNC-2000 oops

PATCH [email protected] 2 participants 2 messages Dec 30, 2007 Memory technology devices (MTD)

Share & tools

[email protected]

author

2007-12-30 13:15 UTC #0 lore raw

while doing some systematical driver load/unload testing i have also come across this one.

searched lkml and found/tried this patch - seems to work fine, at least driver now behaves as expected:
insmod: error inserting 'pnc2000.ko': -1 No such device or address

if it really _works_ is another story, but how can we know ? have searched the net a while for this hardware, but didn`t find _any_ reference that this hardware either exists or existed at all, so it looks that this hardware is very ancient or being used by japanese secret service, only !? :)

Quoted 3 lines

The driver is broken for a long time. I ever ping the writer [email protected] (seems a company's email address), but got no response.

i have found email @crossnet from two people who seem to be involved in software development more or less. have sent them a mail and asked if they know anything about this or could re-establish a contact to the author.

anyway, i vote for the patch being merged to avoid people investigate into this oops or complain here again.

if there is fear, that this breaks the driver even more, what about putting some message into the driver initialization ?

Quoted 4 lines

Quoted 4 lines
Quoted 4 lines
PNC-2000 seems to be very ancient/rare hardware and the driver appears unmaintained for a while. Due to missing testing capabilities, theres no guarantee, that this driver still works. If you own such hardware or if the driver doesnt work for you, please report to [email protected]

regards
roland

List:       linux-kernel
Subject:    [PATCH]: 2.6.24-rc5: mtd: PNC-2000 oops
From:       "Luiz Fernando N. Capitulino" <lcapitulino () mandriva ! com ! br>
Date:       2007-12-14 16:00:26
Message-ID: 20071214140026.621c9007 () mandriva ! com ! br

[Download message RAW]

Em Fri, 14 Dec 2007 10:06:20 +0800
"Dave Young" <[email protected]> escreveu:

Quoted 56 lines

Quoted 54 lines · Luiz Fernando N. Capitulino
Hi there,
If I run:
# modprobe pnc2000
In a machine w/o mtd hardware, I get the following oops:
Photron PNC-2000 flash mapping: 400000 at bf000000
BUG: unable to handle kernel paging request at virtual address bf000000
printing eip: e0bfe05b *pde = 00000000
Oops: 0002 [#1]
Modules linked in: cfi_probe gen_probe pnc2000 mtd chipreg map_funcs nfsd lockd \
nfs_acl auth_rpcgss sunrpc exportfs af_packet snd_seq_dummy snd_seq_oss \
snd_seq_midi_event snd_seq snd_pcm_oss ipv6 snd_mixer_oss binfmt_misc loop \
dm_mirror dm_mod sata_via libata scsi_mod floppy cpufreq_ondemand \
cpufreq_conservative cpufreq_powersave freq_table firewire_ohci firewire_core \
crc_itu_t snd_via82xx gameport snd_ac97_codec ac97_bus snd_pcm snd_timer \
snd_page_alloc snd_mpu401_uart snd_rawmidi snd_seq_device i2c_viapro ehci_hcd snd \
uhci_hcd i2c_core soundcore 8139cp ide_cd cdrom ohci1394 usbcore ieee1394 via_agp \
8139too mii agpgart shpchp pci_hotplug evdev ext3 jbd ide_disk ide_generic \
via82cxxx ide_core
Pid: 4067, comm: modprobe Not tainted (2.6.24-0.rc5.1mdv #1)
EIP: 0060:[<e0bfe05b>] EFLAGS: 00010286 CPU: 0
EIP is at simple_map_write+0x2a/0x34 [map_funcs]
EAX: 00000004 EBX: e0c3f5d0 ECX: bf000000 EDX: 00f000f0
ESI: e0c3f5d0 EDI: 00000002 EBP: dc913dbc ESP: dc913db8
DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
Process modprobe (pid: 4067, ti=dc912000 task=df0409f0 task.ti=dc912000)
Stack: 00000004 dc913e18 e0cc1277 dc913e30 c012b89b 00000000 00000001 00000002
c04a2b7c 00000008 00000000 00000000 c043cde4 00000000 c0357ccb dc913e1c
dc913e1c 00000204 00000000 00000000 e0c00a54 e0cc119d 00000002 00000002
Call Trace:
[<c01054f2>] show_trace_log_lvl+0x1a/0x2f
[<c01055a2>] show_stack_log_lvl+0x9b/0xa3
[<c0105650>] show_registers+0xa6/0x177
[<c0105823>] die+0x102/0x1ee
[<c02cb4d2>] do_page_fault+0x561/0x63f
[<c02c9d1a>] error_code+0x6a/0x70
[<e0cc1277>] cfi_probe_chip+0xda/0x9c7 [cfi_probe]
[<e0c43123>] mtd_do_chip_probe+0x77/0x2d4 [gen_probe]
[<e0cc119b>] cfi_probe+0xd/0xf [cfi_probe]
[<e0c00142>] do_map_probe+0x3d/0x5a [chipreg]
[<e0c4103b>] init_pnc2000+0x3b/0x68 [pnc2000]
[<c0140d10>] sys_init_module+0x12c6/0x1394
[<c0103e42>] sysenter_past_esp+0x6b/0xc9
=======================
Code: c3 55 89 e5 53 89 c3 8b 40 18 83 f8 01 75 07 03 4b 10 88 11 eb 17 83 f8 02 \
75 08 03 4b 10 66 89 11 eb 0a 83 f8 04 75 05 03 4b 10 <89> 11 f0 83 04 24 00 5b \
5d c3 55 89 e5 57 89 d7 8b 55 08 56 89
EIP: [<e0bfe05b>] simple_map_write+0x2a/0x34 [map_funcs] SS:ESP 0068:dc913db8

Quoted 3 lines

Quoted 1 line
--
This issue has been reported several times.

Oh.

Quoted 5 lines

The driver is broken for a long time. I ever ping the writer [email protected] (seems a company's email address), but got no response.
So how about kill (or disable ?) this driver until it is fixed?

Well, looks like there's an ioremap() call missing there.

The following patch does what every driver of that kind does. It
fixes the load OOPS for me but we need someone with the hardware to test whether it works (or some mtd person to review it).

If the patch is wrong I think we should disable the module, yes.

------

mtd: pnc2000: initializaion fix

pnc2000 should map bus memory before trying to access it, also fix wrong struct map_info setup.

Signed-off-by Luiz Fernando N. Capitulino <[email protected]>

drivers/mtd/maps/pnc2000.c +3 -2

Index: linux-2.6.23/drivers/mtd/maps/pnc2000.c
===================================================================
@@ -30,8 +30,7 @@ static struct map_info pnc_map = {
 	.name = "PNC-2000",
 	.size = WINDOW_SIZE,
 	.bankwidth = 4,
-	.phys = 0xFFFFFFFF,
-	.virt = (void __iomem *)WINDOW_ADDR,
+	.phys = WINDOW_ADDR,
 };

@@ -64,16 +63,29 @@ static struct mtd_info *mymtd;
 
 static int __init init_pnc2000(void)
 {
+	int err;
+
 	printk(KERN_NOTICE "Photron PNC-2000 flash mapping: %x at %x\n", WINDOW_SIZE, \

WINDOW_ADDR);
+ pnc_map.virt = ioremap(WINDOW_ADDR, WINDOW_SIZE);
+ if (!pnc_map.virt) {
+ printk("init_pnc2000(): failed to ioremap\n");
+ return -EIO;
+ }
+
simple_map_init(&pnc_map);

mymtd = do_map_probe("cfi_probe", &pnc_map); if (mymtd) {
mymtd->owner = THIS_MODULE;
- return add_mtd_partitions(mymtd, pnc_partitions, 3);
+ err = add_mtd_partitions(mymtd, pnc_partitions, 3);
+ if (err) {
+ printk("init_pnc2000(): add_mtd_partitions failed\n");
+ iounmap(pnc_map.virt);
+ }
}

+ iounmap(pnc_map.virt);
return -ENXIO;
}

@@ -82,6 +94,7 @@ static void __exit cleanup_pnc2000(void)
if (mymtd) {
del_mtd_partitions(mymtd); map_destroy(mymtd);
+ iounmap(pnc_map.virt);
}
}

_______________________________________________________________________
Jetzt neu! Schützen Sie Ihren PC mit McAfee und WEB.DE. 30 Tage kostenlos testen. http://www.pc-sicherheit.web.de/startseite/?mc=022220

Josh Boyer

2007-12-30 14:06 UTC #1 lore raw

Quoted 6 lines · On Sun, 30 Dec 2007 14:15:13 +0100 [email protected] wrote:

while doing some systematical driver load/unload testing i have also come across this one.
searched lkml and found/tried this patch - seems to work fine, at least driver now behaves as expected:
insmod: error inserting 'pnc2000.ko': -1 No such device or address
if it really _works_ is another story, but how can we know ? have searched the net a while for this hardware, but didn`t find _any_ reference that this hardware either exists or existed at all, so it looks that this hardware is very ancient or being used by japanese secret service, only !? :)

[snip]

Quoted 19 lines

(quoted diff) +8 -0

@@ -64,16 +63,29 @@ static struct mtd_info *mymtd;
 
 static int __init init_pnc2000(void)
 {
+	int err;
+
 	printk(KERN_NOTICE "Photron PNC-2000 flash mapping: %x at %x\n", WINDOW_SIZE, \
 WINDOW_ADDR);  
+	pnc_map.virt = ioremap(WINDOW_ADDR, WINDOW_SIZE);
+	if (!pnc_map.virt) {
+		printk("init_pnc2000(): failed to ioremap\n");
+		return -EIO;
+	}
+
 	simple_map_init(&pnc_map);
 
 	mymtd = do_map_probe("cfi_probe", &pnc_map);
 	if (mymtd) {
 		mymtd->owner = THIS_MODULE;

Review static struct mtd_info *mymtd; · line 89

Quoted 10 lines

(quoted diff) +6 -1

-		return add_mtd_partitions(mymtd, pnc_partitions, 3);
+		err = add_mtd_partitions(mymtd, pnc_partitions, 3);
+		if (err) {
+			printk("init_pnc2000(): add_mtd_partitions failed\n");
+			iounmap(pnc_map.virt);
+		}
 	}
 
+	iounmap(pnc_map.virt);
 	return -ENXIO;

Sorry, but no. This bit here is bogus. It will now iounmap and return an error even if nothing went wrong.

josh