Skip to contentSkip to conversation

usb: ehci: add struct entry for broadcom insnreg

PATCH In review Arnd Bergmann 1 participant 2 messages May 27, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 usb: ehci: add struct entry for broadcom insnreg

The newly added broadcom ehci driver triggered another array overflow warning after I had fixed up a bunch of others. In this case, the driver intentionally abuses the port_status[] array to access a register that does not have an official name:

drivers/usb/host/ehci-brcm.c:113:33: error: array index 16 is past the end of the array (which contains 15 elements) [-Werror,-Warray-bounds]

        ehci_writel(ehci, 0x00800040, &ehci->regs->port_status[0x10]);
                                       ^                       ~~~~

include/linux/usb/ehci_def.h:131:3: note: array 'port_status' declared here
u32 port_status[15]; /* up to N_PORTS */
^

There is already a hack for Intel specific registers at the same location, so extend that hack to also cover the Broadcom registers.

I'm a little confused about the register offset, as the code comment says @0x90 while the actual offset seems to be at offset 0x84, please confirm that the behavior is still correct.

Fixes 9df231511bd6 ("usb: ehci: Add new EHCI driver for Broadcom STB SoC's")
Fixes 88aa39691cea ("usb: ehci: avoid gcc-10 zero-length-bounds warning")
---
 drivers/usb/host/ehci-brcm.c | 13 ++++++-------
 include/linux/usb/ehci_def.h |  1 +
 2 files changed, 7 insertions(+), 7 deletions(-)
drivers/usb/host/ehci-brcm.c +6 -7
index 3e0ebe8cc649..bafb3f4db170 100644
@@ -108,10 +108,10 @@ static int ehci_brcm_reset(struct usb_hcd *hcd)
 	/*
 	 * SWLINUX-1705: Avoid OUT packet underflows during high memory
 	 *   bus usage
-	 * port_status[0x0f] = Broadcom-proprietary USB_EHCI_INSNREG00 @ 0x90
+	 * Broadcom-proprietary USB_EHCI_INSNREG00 @ 0x90
 	 */
-	ehci_writel(ehci, 0x00800040, &ehci->regs->port_status[0x10]);
-	ehci_writel(ehci, 0x00000001, &ehci->regs->port_status[0x12]);
+	ehci_writel(ehci, 0x00800040, &ehci->regs->brcm_insnreg[0]);
+	ehci_writel(ehci, 0x00000001, &ehci->regs->brcm_insnreg[2]);
 
 	return ehci_setup(hcd);
 }
@@ -223,11 +223,10 @@ static int __maybe_unused ehci_brcm_resume(struct device *dev)
 	/*
 	 * SWLINUX-1705: Avoid OUT packet underflows during high memory
 	 *   bus usage
-	 * port_status[0x0f] = Broadcom-proprietary USB_EHCI_INSNREG00
-	 * @ 0x90
+	 * Broadcom-proprietary USB_EHCI_INSNREG00 @ 0x90
 	 */
-	ehci_writel(ehci, 0x00800040, &ehci->regs->port_status[0x10]);
-	ehci_writel(ehci, 0x00000001, &ehci->regs->port_status[0x12]);
+	ehci_writel(ehci, 0x00800040, &ehci->regs->brcm_insnreg[0]);
+	ehci_writel(ehci, 0x00000001, &ehci->regs->brcm_insnreg[2]);
 
 	ehci_resume(hcd, false);
 
include/linux/usb/ehci_def.h +1 -0
index 99481bbcc8f7..cf4bfbbdc8a3 100644
@@ -190,6 +190,7 @@ struct ehci_regs {
 #define HOSTPC_PHCD	(1<<22)		/* Phy clock disable */
 #define HOSTPC_PSPD	(3<<25)		/* Port speed detection */
 
+		u32		brcm_insnreg[3]; /* Broadcom specific */
 		u32		reserved6[17];
 	};
 
Signature · 1 line
2.26.2
#1 lore raw
Quoted 4 lines · Arnd Bergmann
Fixes 9df231511bd6 ("usb: ehci: Add new EHCI driver for Broadcom STB SoC's")
Fixes 88aa39691cea ("usb: ehci: avoid gcc-10 zero-length-bounds warning")

Please disregard this submission, I just noticed that the 88aa39691cea commit I referenced is one in my local tree, not in linux-next. I'll make sure they are both good and then resubmit them as a series or a combined patch.

Arnd