linux, branch v3.2.21

Linux 3.2.21

2012-06-19T22:18:30Z

drm/radeon: add some additional 6xx/7xx/EG register init

2012-06-19T22:18:29Z

commit b866d1334ba2d544bc575d75357dea6bdcdc7f46 upstream. - SMX_SAR_CTL0 needs to be programmed correctly to prevent problems with memory exports in certain cases. - VC_ENHANCE needs to be initialized on 6xx/7xx. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie Signed-off-by: Ben Hutchings

swap: fix shmem swapping when more than 8 areas

2012-06-19T22:18:28Z

commit 9b15b817f3d62409290fd56fe3cbb076a931bb0a upstream. Minchan Kim reports that when a system has many swap areas, and tmpfs swaps out to the ninth or more, shmem_getpage_gfp()'s attempts to read back the page cannot locate it, and the read fails with -ENOMEM. Whoops. Yes, I blindly followed read_swap_header()'s pte_to_swp_entry( swp_entry_to_pte()) technique for determining maximum usable swap offset, without stopping to realize that that actually depends upon the pte swap encoding shifting swap offset to the higher bits and truncating it there. Whereas our radix_tree swap encoding leaves offset in the lower bits: it's swap "type" (that is, index of swap area) that was truncated. Fix it by reducing the SWP_TYPE_SHIFT() in swapops.h, and removing the broken radix_to_swp_entry(swp_to_radix_entry()) from read_swap_header(). This does not reduce the usable size of a swap area any further, it leaves it as claimed when making the original commit: no change from 3.0 on x86_64, nor on i386 without PAE; but 3.0's 512GB is reduced to 128GB per swapfile on i386 with PAE. It's not a change I would have risked five years ago, but with x86_64 supported for ten years, I believe it's appropriate now. Hmm, and what if some architecture implements its swap pte with offset encoded below type? That would equally break the maximum usable swap offset check. Happily, they all follow the same tradition of encoding offset above type, but I'll prepare a check on that for next. Reported-and-Reviewed-and-Tested-by: Minchan Kim Signed-off-by: Hugh Dickins Signed-off-by: Linus Torvalds Signed-off-by: Ben Hutchings

USB: fix gathering of interface associations

2012-06-19T22:18:28Z

commit b3a3dd074f7053ef824ad077e5331b52220ceba1 upstream. TEAC's UD-H01 (and probably other devices) have a gap in the interface number allocation of their descriptors: Configuration Descriptor: bLength 9 bDescriptorType 2 wTotalLength 220 bNumInterfaces 3 [...] Interface Descriptor: bLength 9 bDescriptorType 4 bInterfaceNumber 0 bAlternateSetting 0 [...] Interface Association: bLength 8 bDescriptorType 11 bFirstInterface 2 bInterfaceCount 2 bFunctionClass 1 Audio bFunctionSubClass 0 bFunctionProtocol 32 iFunction 4 Interface Descriptor: bLength 9 bDescriptorType 4 bInterfaceNumber 2 bAlternateSetting 0 [...] Once a configuration is selected, usb_set_configuration() walks the known interfaces of a given configuration and calls find_iad() on each of them to set the interface association pointer the interface is included in. The problem here is that the loop variable is taken for the interface number in the comparison logic that gathers the association. Which is fine as long as the descriptors are sane. In the case above, however, the logic gets out of sync and the interface association fields of all interfaces beyond the interface number gap are wrong. Fix this by passing the interface's bInterfaceNumber to find_iad() instead. Signed-off-by: Daniel Mack Reported-by: bEN Reported-by: Ivan Perrone Tested-by: ivan perrone Signed-off-by: Greg Kroah-Hartman Signed-off-by: Ben Hutchings

usb: cdc-acm: fix devices not unthrottled on open

2012-06-19T22:18:28Z

commit 6c4707f3f8c44ec18282e1c014c80e1c257042f9 upstream. Currently CDC-ACM devices stay throttled when their TTY is closed while throttled, stalling further communication attempts after the next open. Unthrottling during open/activate got lost starting with kernel 3.0.0 and this patch reintroduces it. Signed-off-by: Otto Meta Acked-by: Johan Hovold Signed-off-by: Greg Kroah-Hartman Signed-off-by: Ben Hutchings

USB: fix PS3 EHCI systems

2012-06-19T22:18:27Z

commit 4f7a67e2dd49fbfba002c453bc24bf00e701cc71 upstream. After commit aaa0ef289afe9186f81e2340114ea413eef0492a "PS3 EHCI QH read work-around", Terratec Grabby (em28xx) stopped working with AMD Geode LX 800 (USB controller AMD CS5536). Since this is a PS3 only fix, the following patch adds a conditional block around it. Signed-off-by: Ricardo Martins Acked-by: Alan Stern Signed-off-by: Greg Kroah-Hartman Signed-off-by: Ben Hutchings

usb: PS3 EHCI QH read work-around

2012-06-19T22:18:27Z

commit aaa0ef289afe9186f81e2340114ea413eef0492a upstream. PS3 EHCI HC errata fix 244. The SCC EHCI HC will not correctly perform QH reads that occur near or span a micro-frame boundry. This is due to a problem in the Nak Count Reload Control logic (EHCI Specification 1.0 Section 4.9.1). The work-around for this problem is for the HC driver to set I=1 (inactive) for QHs with H=1 (list head). Signed-off-by: Geoff Levand Acked-by: Alan Stern Signed-off-by: Ben Hutchings

xHCI: Increase the timeout for controller save/restore state operation

2012-06-19T22:18:26Z

commit 622eb783fe6ff4c1baa47db16c3a5db97f9e6e50 upstream. When system software decides to power down the xHC with the intent of resuming operation at a later time, it will ask xHC to save the internal state and restore it when resume to correctly recover from a power event. Two bits are used to enable this operation: Save State and Restore State. xHCI spec 4.23.2 says software should "Set the Controller Save/Restore State flag in the USBCMD register and wait for the Save/Restore State Status flag in the USBSTS register to transition to '0'". However, it does not define how long software should wait for the SSS/RSS bit to transition to 0. Currently the timeout is set to 1ms. There is bug report (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1002697) indicates that the timeout is too short for ASMedia ASM1042 host controller to save/restore the state successfully. Increase the timeout to 10ms helps to resolve the issue. This patch should be backported to stable kernels as old as 2.6.37, that contain the commit 5535b1d5f8885695c6ded783c692e3c0d0eda8ca "USB: xHCI: PCI power management implementation" Signed-off-by: Andiry Xu Signed-off-by: Sarah Sharp Cc: Ming Lei Signed-off-by: Ben Hutchings

xhci: Don't free endpoints in xhci_mem_cleanup()

2012-06-19T22:18:26Z

commit 32f1d2c536d0c26c5814cb0e6a0606c42d02fac1 upstream. This patch fixes a few issues introduced in the recent fix [f8a9e72d: USB: fix resource leak in xhci power loss path] - The endpoints listed in bw table are just links and each entry is an array member of dev->eps[]. But the commit above adds a kfree() call to these instances, and thus it results in memory corruption. - It clears only the first entry of rh_bw[], but there can be multiple ports. - It'd be safer to clear the list_head of ep as well, not only removing from the list, as it's checked in xhci_discover_or_reset_device(). This patch should be backported to kernels as old as 3.2, that contain the commit 839c817ce67178ca3c7c7ad534c571bba1e69ebe "xhci: Store information about roothubs and TTs." Signed-off-by: Takashi Iwai Signed-off-by: Sarah Sharp Reviewed-by: Oliver Neukum Signed-off-by: Ben Hutchings

xhci: Fix invalid loop check in xhci_free_tt_info()

2012-06-19T22:18:25Z

commit 46ed8f00d8982e49f8fe2c1a9cea192f640cb3ba upstream. xhci_free_tt_info() may access the invalid memory when it removes the last entry but the list is not empty. Then tt_next reaches to the list head but it still tries to check the tt_info of that entry. This patch fixes the bug and cleans up the messy code by rewriting with a simple list_for_each_entry_safe(). This patch should be backported to kernels as old as 3.2, that contain the commit 839c817ce67178ca3c7c7ad534c571bba1e69ebe "xhci: Store information about roothubs and TTs." Signed-off-by: Takashi Iwai Signed-off-by: Sarah Sharp Reviewed-by: Oliver Neukum Signed-off-by: Ben Hutchings