Re: [bisected] NULL pointer dereference in acpi_gpiochip_add (on modprobe ssb)

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Mon, Mar 24, 2014 at 04:49:44PM +0100, Sabrina Dubroca wrote:
> Hi,
> 
> With next-20140324, I get the BUG below when I modprobe ssb.
> I bisected it to aa92b6f689ac
> "gpio / ACPI: Allocate ACPI specific data directly in acpi_gpiochip_add()"
> 
> The device that needs ssb is:
> 0c:00.0 Network controller [0280]: Broadcom Corporation BCM4321 802.11a/b/g/n [14e4:4328] (rev 03)
>         Subsystem: Dell Wireless 1500 Draft 802.11n WLAN Mini-card [1028:000a]
>         Kernel driver in use: b43-pci-bridge
>         Kernel modules: ssb
> 
> 
> [   92.693606] ssb: Found chip with id 0x4321, rev 0x03 and package 0x00
> [   92.693649] ssb: Core 0 found: ChipCommon (cc 0x800, rev 0x13, vendor 0x4243)
> [   92.693675] ssb: Core 1 found: IEEE 802.11 (cc 0x812, rev 0x0C, vendor 0x4243)
> [   92.693699] ssb: Core 2 found: PCI-E (cc 0x820, rev 0x04, vendor 0x4243)
> [   92.693723] ssb: Core 3 found: PCI (cc 0x804, rev 0x0D, vendor 0x4243)
> [   92.693746] ssb: Core 4 found: USB 1.1 Host (cc 0x817, rev 0x04, vendor 0x4243)
> [   92.753554] BUG: unable to handle kernel NULL pointer dereference at 00000138
> [   92.753760] IP: [<c126c2b3>] acpi_gpiochip_add+0x13/0x190
> [   92.753901] *pde = 00000000 
> [   92.753986] Oops: 0000 [#1] PREEMPT SMP 
> [   92.754125] Modules linked in: ssb(+) mmc_core netconsole nouveau mxm_wmi i2c_algo_bit drm_kms_helper ttm drm joydev mousedev tg3 coretemp kvm_intel ptp pcmcia kvm pps_core libphy dell_laptop gpio_ich rfkill yenta_socket pcmcia_rsrc intel_agp intel_gtt iTCO_wdt iTCO_vendor_support dell_wmi sparse_keymap pcmcia_core evdev agpgart dcdbas snd_hda_codec_idt snd_hda_codec_generic microcode psmouse pcspkr i2c_i801 i2c_core serio_raw lpc_ich mfd_core acpi_cpufreq ac battery thermal button wmi snd_hda_intel snd_hda_controller snd_hda_codec snd_hwdep snd_pcm video snd_timer shpchp processor snd soundcore nfs lockd sunrpc ext4 crc16 mbcache jbd2 sd_mod sr_mod cdrom ata_generic pata_acpi ata_piix libata scsi_mod firewire_ohci firewire_core crc_itu_t uhci_hcd ehci_pci ehci_hcd usbcore usb_common
> [   92.756833] CPU: 0 PID: 512 Comm: modprobe Tainted: G        W     3.14.0-rc7-next-20140324-t1 #24
> [   92.756833] Hardware name: Dell Inc. Latitude D830                   /0UY141, BIOS A02 06/07/2007
> [   92.756833] task: f5799900 ti: f543e000 task.ti: f543e000
> [   92.756833] EIP: 0060:[<c126c2b3>] EFLAGS: 00010282 CPU: 0
> [   92.756833] EIP is at acpi_gpiochip_add+0x13/0x190
> [   92.756833] EAX: 00000000 EBX: f57824c4 ECX: 00000000 EDX: 00000000
> [   92.756833] ESI: f57824c4 EDI: 00000010 EBP: f543fc54 ESP: f543fc40
> [   92.756833]  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
> [   92.756833] CR0: 8005003b CR2: 00000138 CR3: 355f8000 CR4: 000007d0

To me looks like chip->dev is NULL. My understanding is that the GPIO core
wants to have it non-NULL.

Can you try if following patch helps?

diff --git a/drivers/ssb/driver_gpio.c b/drivers/ssb/driver_gpio.c
index ba350d2035c0..3cbf74a47851 100644
--- a/drivers/ssb/driver_gpio.c
+++ b/drivers/ssb/driver_gpio.c
@@ -234,6 +234,8 @@ static int ssb_gpio_chipco_init(struct ssb_bus *bus)
 	chip->to_irq		= ssb_gpio_to_irq;
 #endif
 	chip->ngpio		= 16;
+	chip->dev		= bus->chipco.dev;
+
 	/* There is just one SoC in one device and its GPIO addresses should be
 	 * deterministic to address them more easily. The other buses could get
 	 * a random base number. */