Patch series "userfaultfd non-cooperative further update for 4.11 merge
window".
Unfortunately I noticed one relevant bug in userfaultfd_exit while doing
more testing. I've been doing testing before and this was also tested
by kbuild bot and exercised by the selftest, but this bug never
reproduced before.
I dropped userfaultfd_exit as result. I dropped it because of
implementation difficulty in receiving signals in __mmput and because I
think -ENOSPC as result from the background UFFDIO_COPY should be enough
already.
Before I decided to remove userfaultfd_exit, I noticed userfaultfd_exit
wasn't exercised by the selftest and when I tried to exercise it, after
moving it to a more correct place in __mmput where it would make more
sense and where the vma list is stable, it resulted in the
event_wait_completion in D state. So then I added the second patch to
be sure even if we call userfaultfd_event_wait_completion too late
during task exit(), we won't risk to generate tasks in D state. The
same check exists in handle_userfault() for the same reason, except it
makes a difference there, while here is just a robustness check and it's
run under WARN_ON_ONCE.
While looking at the userfaultfd_event_wait_completion() function I
looked back at its callers too while at it and I think it's not ok to
stop executing dup_fctx on the fcs list because we relay on
userfaultfd_event_wait_completion to execute
userfaultfd_ctx_put(fctx->orig) which is paired against
userfaultfd_ctx_get(fctx->orig) in dup_userfault just before
list_add(fcs). This change only takes care of fctx->orig but this area
also needs further review looking for similar problems in fctx->new.
The only patch that is urgent is the first because it's an use after
free during a SMP race condition that affects all processes if
CONFIG_USERFAULTFD=y. Very hard to reproduce though and probably
impossible without SLUB poisoning enabled.
This patch (of 3):
I once reproduced this oops with the userfaultfd selftest, it's not
easily reproducible and it requires SLUB poisoning to reproduce.
In the debugger I located the "mm" pointer in the stack and walking
mm->mmap->vm_next through the end shows the vma->vm_next list is fully
consistent and it is null terminated list as expected. So this has to
be an SMP race condition where userfaultfd_exit was running while the
vma list was being modified by another CPU.
When userfaultfd_exit() run one of the ->vm_next pointers pointed to
SLAB_POISON (RBX is the vma pointer and is 0x6b6b..).
The reason is that it's not running in __mmput but while there are still
other threads running and it's not holding the mmap_sem (it can't as it
has to wait the even to be received by the manager). So this is an use
after free that was happening for all processes.
One more implementation problem aside from the race condition:
userfaultfd_exit has really to check a flag in mm->flags before walking
the vma or it's going to slowdown the exit() path for regular tasks.
One more implementation problem: at that point signals can't be
delivered so it would also create a task in D state if the manager
doesn't read the event.
The major design issue: it overall looks superfluous as the manager can
check for -ENOSPC in the background transfer:
if (mmget_not_zero(ctx->mm)) {
[..]
} else {
return -ENOSPC;
}
It's safer to roll it back and re-introduce it later if at all.
Dan Williams [Fri, 10 Mar 2017 00:16:45 +0000 (16:16 -0800)]
x86, mm: unify exit paths in gup_pte_range()
All exit paths from gup_pte_range() require pte_unmap() of the original
pte page before returning. Refactor the code to have a single exit
point to do the unmap.
This mirrors the flow of the generic gup_pte_range() in mm/gup.c.
Dan Williams [Fri, 10 Mar 2017 00:16:42 +0000 (16:16 -0800)]
x86, mm: fix gup_pte_range() vs DAX mappings
gup_pte_range() fails to check pte_allows_gup() before translating a DAX
pte entry, pte_devmap(), to a page. This allows writes to read-only
mappings, and bypasses the DAX cacheline dirty tracking due to missed
'mkwrite' faults. The gup_huge_pmd() path and the gup_huge_pud() path
correctly check pte_allows_gup() before checking for _devmap() entries.
power/mm: update pte_write and pte_wrprotect to handle savedwrite
We use pte_write() to check whethwer the pte entry is writable. This is
mostly used to later mark the pte read only if it is writable. The other
use of pte_write() is to check whether the pte_entry is writable so that
hardware page table entry can be marked accordingly. This is used in kvm
where we look at qemu page table entry and update hardware hash page table
for the guest with correct write enable bit.
With the above, for the first usage we should also check the savedwrite
bit so that we can correctly clear the savedwite bit. For the later, we
add a new variant __pte_write().
With this we can revert write_protect_page part of 595cd8f256d2 ("mm/ksm:
handle protnone saved writes when making page write protect"). But I left
it as it is as an example code for savedwrite check.
We need to mark pages of parent process read only on fork. Numa fault
pte needs a protnone ptes variant with saved write flag set. On fork we
need to make sure we remove the saved write bit. Instead of adding the
protnone check in the caller update ptep_set_wrprotect variants to clear
savedwrite bit.
Without this we see random segfaults in application on fork.
Masahiro Yamada [Fri, 10 Mar 2017 00:16:31 +0000 (16:16 -0800)]
scripts/spelling.txt: add "disble(d)" pattern and fix typo instances
Fix typos and add the following to the scripts/spelling.txt:
disble||disable
disbled||disabled
I kept the TSL2563_INT_DISBLED in /drivers/iio/light/tsl2563.c
untouched. The macro is not referenced at all, but this commit is
touching only comment blocks just in case.
__do_fault assumes vmf->page has been initialized and is valid if
VM_FAULT_NOPAGE is not returned by vma->vm_ops->fault(vma, vmf).
handle_userfault() in turn should return VM_FAULT_NOPAGE if it doesn't
return VM_FAULT_SIGBUS or VM_FAULT_RETRY (the other two possibilities).
This VM_FAULT_NOPAGE case is only invoked when signal are pending and it
didn't matter for anonymous memory before. It only started to matter
since shmem was introduced. hugetlbfs also takes a different path and
doesn't exercise __do_fault.
Linus Torvalds [Fri, 10 Mar 2017 00:30:37 +0000 (16:30 -0800)]
Merge tag 'pm-4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fixes from Rafael Wysocki:
"These fix several issues in the intel_pstate driver and one issue in
the schedutil cpufreq governor, clean up that governor a bit and hook
up existing code for disabling cpufreq to a new kernel command line
option.
Specifics:
- Three fixes for intel_pstate problems related to the passive mode
(in which it acts as a regular cpufreq scaling driver), two for the
handling of global P-state limits and one for the handling of the
cpu_frequency tracepoint in that mode (Rafael Wysocki).
- Three fixes for the handling of P-state limits in intel_pstate in
the active mode (Rafael Wysocki).
- Introduction of a new cpufreq.off=1 kernel command line argument
that will disable cpufreq entirely if passed to the kernel and is
simply hooked up to the existing code used by Xen (Len Brown).
- Fix for the schedutil cpufreq governor to prevent it from using
stale raw frequency values in configurations with mutiple CPUs
sharing one policy object and a cleanup for it reducing its
overhead slightly (Viresh Kumar)"
* tag 'pm-4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: intel_pstate: Do not reinit performance limits in ->setpolicy
cpufreq: intel_pstate: Fix intel_pstate_verify_policy()
cpufreq: intel_pstate: Fix global settings in active mode
cpufreq: Add the "cpufreq.off=1" cmdline option
cpufreq: schedutil: Pass sg_policy to get_next_freq()
cpufreq: schedutil: move cached_raw_freq to struct sugov_policy
cpufreq: intel_pstate: Avoid triggering cpu_frequency tracepoint unnecessarily
cpufreq: intel_pstate: Fix intel_cpufreq_verify_policy()
cpufreq: intel_pstate: Do not use performance_limits in passive mode
Linus Torvalds [Fri, 10 Mar 2017 00:02:04 +0000 (16:02 -0800)]
Merge tag 'pci-v4.11-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Pull PCI fixes from Bjorn Helgaas:
"PCI fixes:
- fix NULL pointer dereference in Exynos driver
- fix NULL pointer dereference in ASPM with pre-1.1 PCIe devices
- blacklist QLogic ISP2722 to prevent panics while reading VPD"
* tag 'pci-v4.11-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
PCI/ASPM: Always set link->downstream to avoid NULL dereference on remove
PCI: Prevent VPD access for QLogic ISP2722
PCI: exynos: Initialize elbi_base even when using PHY framework
Linus Torvalds [Thu, 9 Mar 2017 23:53:25 +0000 (15:53 -0800)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
"Sending this a bit sooner than I otherwise would have, as a fix in the
merge window had some unfortunate issues and side effects for some
folks.
This contains:
- Fixes from Jan for the bdi registration/unregistration. These have
been tested by the various parties reporting issues, and should be
solid at this point.
- Also from Jan, fix for axonram gendisk registration.
- A stable fix for zram from Johannes.
- A small series from Ming, fixing up some long standing issues with
blk-mq hardware queue kobject initialization and registration.
- A fix for sed opal from Jon, fixing a nonsensical range check and
some set-but-not-used variables.
- A fix from Neil for a long standing deadlock issue for stacking
device drivers. With this in place, dm/md don't have to work around
the issue anymore, and can be properly fixed up"
* 'for-linus' of git://git.kernel.dk/linux-block:
axonram: Fix gendisk handling
blk: improve order of bio handling in generic_make_request()
Revert "scsi, block: fix duplicate bdi name registration crashes"
block: Make del_gendisk() safer for disks without queues
bdi: Fix use-after-free in wb_congested_put()
block: Allow bdi re-registration
block/sed: Fix opal user range check and unused variables
zram: set physical queue limits to avoid array out of bounds accesses
blk-mq: free hctx->cpumask in release handler of hctx's kobject
blk-mq: make lifetime consistent between hctx and its kobject
blk-mq: make lifetime consitent between q/ctx and its kobject
blk-mq: initialize mq kobjects in blk_mq_init_allocated_queue()
Linus Torvalds [Thu, 9 Mar 2017 23:50:56 +0000 (15:50 -0800)]
Merge tag 'media/v4.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
Pull media fixes from Mauro Carvalho Chehab:
"Media regression fixes:
- serial_ir: fix a Kernel crash during boot on Kernel 4.11-rc1, due
to an IRQ code called too early
- other IR regression fixes at lirc and at the raw IR decoding
- a deadlock fix at the RC nuvoton driver
- fix another issue with DMA on stack at dw2102 driver
There's an extra patch there that change a driver interface for the
SoC VSP1 driver, with is shared between the DRM and V4L2 driver. The
patch itself is trivial, and was acked by David Arlie"
* tag 'media/v4.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
[media] v4l: vsp1: Adapt vsp1_du_setup_lif() interface to use a structure
[media] dw2102: don't do DMA on stack
[media] rc: protocol is not set on register for raw IR devices
[media] rc: raw decoder for keymap protocol is not loaded on register
[media] rc: nuvoton: fix deadlock in nvt_write_wakeup_codes
[media] lirc: fix dead lock between open and wakeup_filter
[media] serial_ir: ensure we're ready to receive interrupts
when all map elements are pre-allocated one cpu can delete and reuse htab_elem
while another cpu is still walking the hlist. In such case the lookup may
miss the element. Convert hlist to hlist_nulls to avoid such scenario.
When bucket lock is taken there is no need to take such precautions,
so only convert map_lookup and map_get_next to nulls.
The race window is extremely small and only reproducible with explicit
udelay() inside lookup_nulls_elem_raw()
Similar to hlist add hlist_nulls_for_each_entry_safe() and
hlist_nulls_entry_safe() helpers.
when htab_elem is removed from the bucket list the htab_elem.hash_node.next
field should not be overridden too early otherwise we have a tiny race window
between lookup and delete.
The bug was discovered by manual code analysis and reproducible
only with explicit udelay() in lookup_elem_raw().
Replace MAX_ADDR_LEN with its numeric value to fix the following
linux/packet_diag.h userspace compilation error:
/usr/include/linux/packet_diag.h:67:17: error: 'MAX_ADDR_LEN' undeclared here (not in a function)
__u8 pdmc_addr[MAX_ADDR_LEN];
This is not the first case in the UAPI where the numeric value
of MAX_ADDR_LEN is used instead of symbolic one, uapi/linux/if_link.h
already does the same:
Paolo Abeni [Tue, 7 Mar 2017 17:33:31 +0000 (18:33 +0100)]
net/tunnel: set inner protocol in network gro hooks
The gso code of several tunnels type (gre and udp tunnels)
takes for granted that the skb->inner_protocol is properly
initialized and drops the packet elsewhere.
On the forwarding path no one is initializing such field,
so gro encapsulated packets are dropped on forward.
Since commit 38720352412a ("gre: Use inner_proto to obtain
inner header protocol"), this can be reproduced when the
encapsulated packets use gre as the tunneling protocol.
The issue happens also with vxlan and geneve tunnels since
commit 8bce6d7d0d1e ("udp: Generalize skb_udp_segment"), if the
forwarding host's ingress nic has h/w offload for such tunnel
and a vxlan/geneve device is configured on top of it, regardless
of the configured peer address and vni.
To address the issue, this change initialize the inner_protocol
field for encapsulated packets in both ipv4 and ipv6 gro complete
callbacks.
Fixes: 38720352412a ("gre: Use inner_proto to obtain inner header protocol") Fixes: 8bce6d7d0d1e ("udp: Generalize skb_udp_segment") Signed-off-by: Paolo Abeni <[email protected]> Acked-by: Alexander Duyck <[email protected]> Signed-off-by: David S. Miller <[email protected]>
In qed_ll2_start_ooo() the ll2_info variable is uninitialized and then
passed to qed_ll2_acquire_connection() where it is copied into a new
memory space.
This shouldn't cause any issue as long as non of the copied memory is
every read.
But the potential for a bug being introduced by reading this memory
is real.
Detected by CoverityScan, CID#1399632 ("Uninitialized scalar variable")
This patch set fixes multiple issues such as IOMMU
translation faults when kernel is booted with IOMMU enabled
on host, incorrect MAC ID reading from ACPI tables and IPv6
UDP packet drop due to failure of checksum validation.
Changes from v1:
- As suggested by David Miller, got rid of conditional
calling of DMA map/unmap APIs. Also updated commit message
in 'IOMMU translation faults' patch.
====================
Sunil Goutham [Tue, 7 Mar 2017 12:39:10 +0000 (18:09 +0530)]
net: thunderx: Fix invalid mac addresses for node1 interfaces
When booted with ACPI, random mac addresses are being
assigned to node1 interfaces due to mismatch of bgx_id
in BGX driver and ACPI tables.
This patch fixes this issue by setting maximum BGX devices
per node based on platform/soc instead of a macro. This
change will set the bgx_id appropriately.
Sunil Goutham [Tue, 7 Mar 2017 12:39:09 +0000 (18:09 +0530)]
net: thunderx: Fix LMAC mode debug prints for QSGMII mode
When BGX/LMACs are in QSGMII mode, for some LMACs, mode info is
not being printed. This patch will fix that. With changes already
done to not do any sort of serdes 2 lane mapping config calculation
in kernel driver, we can get rid of this logic.
Sunil Goutham [Tue, 7 Mar 2017 12:39:08 +0000 (18:09 +0530)]
net: thunderx: Fix IOMMU translation faults
ACPI support has been added to ARM IOMMU driver in 4.10 kernel
and that has resulted in VNIC interfaces throwing translation
faults when kernel is booted with ACPI as driver was not using
DMA API. This patch fixes the issue by using DMA API which inturn
will create translation tables when IOMMU is enabled.
Also VNIC doesn't have a seperate receive buffer ring per receive
queue, so there is no 1:1 descriptor index matching between CQE_RX
and the index in buffer ring from where a buffer has been used for
DMA'ing. Unlike other NICs, here it's not possible to maintain dma
address to virt address mappings within the driver. This leaves us
no other choice but to use IOMMU's IOVA address conversion API to
get buffer's virtual address which can be given to network stack
for processing.
Zhu Yanjun [Tue, 7 Mar 2017 07:48:36 +0000 (02:48 -0500)]
rds: ib: add error handle
In the function rds_ib_setup_qp, the error handle is missing. When some
error occurs, it is possible that memory leak occurs. As such, error
handle is added.
VSR Burru [Tue, 7 Mar 2017 02:45:59 +0000 (18:45 -0800)]
liquidio: improve UDP TX performance
Improve UDP TX performance by:
* reducing the ring size from 2K to 512
* replacing the numerous streaming DMA allocations for info buffers and
gather lists with one large consistent DMA allocation per ring
BQL is not effective here. We reduced the ring size because there is heavy
overhead with dma_map_single every so often. With iommu=on, dma_map_single
in PF Tx data path was taking longer time (~700usec) for every ~250
packets. Debugged intel_iommu code, and found that PF driver is utilizing
too many static IO virtual address mapping entries (for gather list entries
and info buffers): about 100K entries for two PF's each using 8 rings.
Also, finding an empty entry (in rbtree of device domain's iova mapping in
kernel) during Tx path becomes a bottleneck every so often; the loop to
find the empty entry goes through over 40K iterations; this is too costly
and was the major overhead. Overhead is low when this loop quits quickly.
David Ahern [Mon, 6 Mar 2017 23:57:31 +0000 (15:57 -0800)]
net: ipv6: Remove redundant RTA_OIF in multipath routes
Dinesh reported that RTA_MULTIPATH nexthops are 8-bytes larger with IPv6
than IPv4. The recent refactoring for multipath support in netlink
messages does discriminate between non-multipath which needs the OIF
and multipath which adds a rtnexthop struct for each hop making the
RTA_OIF attribute redundant. Resolve by adding a flag to the info
function to skip the oif for multipath.
Fixes: beb1afac518d ("net: ipv6: Add support to dump multipath routes
via RTA_MULTIPATH attribute") Reported-by: Dinesh Dutt <[email protected]> Signed-off-by: David Ahern <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Linus Torvalds [Thu, 9 Mar 2017 20:23:30 +0000 (12:23 -0800)]
Merge tag 'for-linus-4.11-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip
Pull xen fix and cleanup from Juergen Gross:
"This contains one fix for MSIX handling under Xen and a trivial
cleanup patch"
* tag 'for-linus-4.11-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xenbus: Remove duplicate inclusion of linux/init.h
xen: do not re-use pirq number cached in pci device msi msg data
arch, mm: convert all architectures to use 5level-fixup.h
If an architecture uses 4level-fixup.h we don't need to do anything as
it includes 5level-fixup.h.
If an architecture uses pgtable-nop*d.h, define __ARCH_USE_5LEVEL_HACK
before inclusion of the header. It makes asm-generic code to use
5level-fixup.h.
If an architecture has 4-level paging or folds levels on its own,
include 5level-fixup.h directly.
We are going to introduce <asm-generic/pgtable-nop4d.h> to provide
abstraction for properly (in opposite to 5level-fixup.h hack) folded
p4d level. The new header will be included from pgtable-nopud.h.
If an architecture uses <asm-generic/nop*d.h>, we cannot use
5level-fixup.h directly to quickly convert the architecture to 5-level
paging as it would conflict with pgtable-nop4d.h.
With this patch an architecture can define __ARCH_USE_5LEVEL_HACK before
inclusion <asm-genenric/nop*d.h> to use 5level-fixup.h.
We are going to switch core MM to 5-level paging abstraction.
This is preparation step which adds <asm-generic/5level-fixup.h>
As with 4level-fixup.h, the new header allows quickly make all
architectures compatible with 5-level paging in core MM.
In long run we would like to switch architectures to properly folded p4d
level by using <asm-generic/pgtable-nop4d.h>, but it requires more
changes to arch-specific code.
Guenter Roeck [Thu, 9 Mar 2017 13:39:37 +0000 (15:39 +0200)]
usb: host: xhci-plat: Fix timeout on removal of hot pluggable xhci controllers
Upstream commit 98d74f9ceaef ("xhci: fix 10 second timeout on removal of
PCI hotpluggable xhci controllers") fixes a problem with hot pluggable PCI
xhci controllers which can result in excessive timeouts, to the point where
the system reports a deadlock.
The same problem is seen with hot pluggable xhci controllers using the
xhci-plat driver, such as the driver used for Type-C ports on rk3399.
Similar to hot-pluggable PCI controllers, the driver for this chip
removes the xhci controller from the system when the Type-C cable is
disconnected.
The solution for PCI devices works just as well for non-PCI devices
and avoids the problem.
Peter Chen [Thu, 9 Mar 2017 13:39:36 +0000 (15:39 +0200)]
usb: host: xhci-dbg: HCIVERSION should be a binary number
According to xHCI spec, HCIVERSION containing a BCD encoding
of the xHCI specification revision number, 0100h corresponds
to xHCI version 1.0. Change "100" as "0x100".
Chunfeng Yun [Thu, 9 Mar 2017 13:39:34 +0000 (15:39 +0200)]
usb: xhci-mtk: check hcc_params after adding primary hcd
hcc_params is set in xhci_gen_setup() called from usb_add_hcd(),
so checks the Maximum Primary Stream Array Size in the hcc_params
register after adding primary hcd.
Yuantian Tang [Thu, 9 Mar 2017 09:13:29 +0000 (17:13 +0800)]
ahci: qoriq: correct the sata ecc setting error
Sata ecc is controlled by only 1 bit which is 24bit in big-endian
in ecc register. So only setting 24bit to disable sata ecc prevents
other bits from being overwritten in ecc register.
Wolfram Sang [Thu, 9 Mar 2017 15:41:48 +0000 (16:41 +0100)]
Revert "i2c: copy device properties when using i2c_register_board_info()"
This reverts commit b0c1e95ab44feaad8831f2c06a3473c974003b49. It
contains a flaw and the next version has more features added which makes
me want to move it to the next cycle.
Wolfram Sang [Thu, 9 Mar 2017 15:32:17 +0000 (16:32 +0100)]
Revert "i2c: add missing of_node_put in i2c_mux_del_adapters"
This reverts commit 02dbfa5e5583523035f05636c614a0eca77f1aab. I grabbed
the wrong version from the list and will pull the proper one from Peter
Rosin's mux tree.
i2c: exynos5: Avoid transaction timeouts due TRANSFER_DONE_AUTO not set
After commit 7999eecb7e56 ("i2c: exynos5: fix arbitration lost handling"),
some I2C transactions are failing because the TRANSFER_DONE_AUTO field is
not set in the I2C_TRANS_STATUS register so the i2c->status value is left
to -EINVAL causing the i2c->msg_complete completion to never be signaled.
For example, when reading the time of an I2C rtc on an Exynos5800 machine:
The Exynos5422 manual states clearly that most I2C_TRANS_STATUS reg bits
(including TRANSFER_DONE_AUTO) are cleared after the register is read. So
reading has side effects and should only be done if HSI2C_INT_I2C was set.
Fixes: 7999eecb7e56 ("i2c: exynos5: fix arbitration lost handling") Signed-off-by: Javier Martinez Canillas <[email protected]> Reviewed-by: Andrzej Hajda <[email protected]> Signed-off-by: Wolfram Sang <[email protected]>
Radim Krčmář [Thu, 9 Mar 2017 14:48:42 +0000 (15:48 +0100)]
Merge tag 'kvm-arm-for-4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm
KVM/ARM updates for v4.11-rc2
vgic updates:
- Honour disabling the ITS
- Don't deadlock when deactivating own interrupts via MMIO
- Correctly expose the lact of IRQ/FIQ bypass on GICv3
I/O virtualization:
- Make KVM_CAP_NR_MEMSLOTS big enough for large guests with
many PCIe devices
General bug fixes:
- Gracefully handle exception generated with syndroms that
the host doesn't understand
- Properly invalidate TLBs on VHE systems
Radim Krčmář [Tue, 7 Mar 2017 16:51:49 +0000 (17:51 +0100)]
KVM: nVMX: do not warn when MSR bitmap address is not backed
Before trying to do nested_get_page() in nested_vmx_merge_msr_bitmap(),
we have already checked that the MSR bitmap address is valid (4k aligned
and within physical limits). SDM doesn't specify what happens if the
there is no memory mapped at the valid address, but Intel CPUs treat the
situation as if the bitmap was configured to trap all MSRs.
KVM already does that by returning false and a correct handling doesn't
need the guest-trigerrable warning that was reported by syzkaller:
(The warning was originally there to catch some possible bugs in nVMX.)
Reported-by: Dmitry Vyukov <[email protected]> Reviewed-by: Jim Mattson <[email protected]>
[Jim Mattson explained the bare metal behavior: "I believe this behavior
would be documented in the chipset data sheet rather than the SDM,
since the chipset returns all 1s for an unclaimed read."] Signed-off-by: Radim Krčmář <[email protected]>
* pm-cpufreq:
cpufreq: intel_pstate: Do not reinit performance limits in ->setpolicy
cpufreq: intel_pstate: Fix intel_pstate_verify_policy()
cpufreq: intel_pstate: Fix global settings in active mode
cpufreq: Add the "cpufreq.off=1" cmdline option
cpufreq: intel_pstate: Avoid triggering cpu_frequency tracepoint unnecessarily
cpufreq: intel_pstate: Fix intel_cpufreq_verify_policy()
cpufreq: intel_pstate: Do not use performance_limits in passive mode
Thomas Gleixner [Thu, 9 Mar 2017 11:06:41 +0000 (12:06 +0100)]
Merge tag 'irq-fixes-4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms into irq/urgent
Pull irqchip/irqdomain updates for 4.11-rc2 from Marc Zyngier
- irqchip/crossbar: Some type tidying up
- irqchip/gicv3-its: Workaround for a Qualcomm erratum
- irqdomain: Compile for for systems that don't use CONFIG_IRQ_DOMAIN
crypto: s5p-sss - Fix spinlock recursion on LRW(AES)
Running TCRYPT with LRW compiled causes spinlock recursion:
testing speed of async lrw(aes) (lrw(ecb-aes-s5p)) encryption
tcrypt: test 0 (256 bit key, 16 byte blocks): 19007 operations in 1 seconds (304112 bytes)
tcrypt: test 1 (256 bit key, 64 byte blocks): 15753 operations in 1 seconds (1008192 bytes)
tcrypt: test 2 (256 bit key, 256 byte blocks): 14293 operations in 1 seconds (3659008 bytes)
tcrypt: test 3 (256 bit key, 1024 byte blocks): 11906 operations in 1 seconds (12191744 bytes)
tcrypt: test 4 (256 bit key, 8192 byte blocks):
BUG: spinlock recursion on CPU#1, irq/84-10830000/89
lock: 0xeea99a68, .magic: dead4ead, .owner: irq/84-10830000/89, .owner_cpu: 1
CPU: 1 PID: 89 Comm: irq/84-10830000 Not tainted 4.11.0-rc1-00001-g897ca6d0800d #559
Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[<c010e1ec>] (unwind_backtrace) from [<c010ae1c>] (show_stack+0x10/0x14)
[<c010ae1c>] (show_stack) from [<c03449c0>] (dump_stack+0x78/0x8c)
[<c03449c0>] (dump_stack) from [<c015de68>] (do_raw_spin_lock+0x11c/0x120)
[<c015de68>] (do_raw_spin_lock) from [<c0720110>] (_raw_spin_lock_irqsave+0x20/0x28)
[<c0720110>] (_raw_spin_lock_irqsave) from [<c0572ca0>] (s5p_aes_crypt+0x2c/0xb4)
[<c0572ca0>] (s5p_aes_crypt) from [<bf1d8aa4>] (do_encrypt+0x78/0xb0 [lrw])
[<bf1d8aa4>] (do_encrypt [lrw]) from [<bf1d8b00>] (encrypt_done+0x24/0x54 [lrw])
[<bf1d8b00>] (encrypt_done [lrw]) from [<c05732a0>] (s5p_aes_complete+0x60/0xcc)
[<c05732a0>] (s5p_aes_complete) from [<c0573440>] (s5p_aes_interrupt+0x134/0x1a0)
[<c0573440>] (s5p_aes_interrupt) from [<c01667c4>] (irq_thread_fn+0x1c/0x54)
[<c01667c4>] (irq_thread_fn) from [<c0166a98>] (irq_thread+0x12c/0x1e0)
[<c0166a98>] (irq_thread) from [<c0136a28>] (kthread+0x108/0x138)
[<c0136a28>] (kthread) from [<c0107778>] (ret_from_fork+0x14/0x3c)
Interrupt handling routine was calling req->base.complete() under
spinlock. In most cases this wasn't fatal but when combined with some
of the cipher modes (like LRW) this caused recursion - starting the new
encryption (s5p_aes_crypt()) while still holding the spinlock from
previous round (s5p_aes_complete()).
Beside that, the s5p_aes_interrupt() error handling path could execute
two completions in case of error for RX and TX blocks.
Rewrite the interrupt handling routine and the completion by:
1. Splitting the operations on scatterlist copies from
s5p_aes_complete() into separate s5p_sg_done(). This still should be
done under lock.
The s5p_aes_complete() now only calls req->base.complete() and it has
to be called outside of lock.
2. Moving the s5p_aes_complete() out of spinlock critical sections.
In interrupt service routine s5p_aes_interrupts(), it appeared in few
places, including error paths inside other functions called from ISR.
This code was not so obvious to read so simplify it by putting the
s5p_aes_complete() only within ISR level.
Merge tag 'usb-serial-4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial into usb-linus
Johan writes:
USB-serial fixes for v4.11-rc2
Here's a fix for a digi_acceleport regression in -rc1, and some fixes
for long-standing issues in three other drivers, including a
NULL-pointer dereference and a couple of information leaks that could be
triggered by a malicious device.
A recent change claimed to fix an off-by-one error in the OOB-port
completion handler, but instead introduced such an error. This could
specifically led to modem-status changes going unnoticed, effectively
breaking TIOCMGET.
Note that the offending commit fixes a loop-condition underflow and is
marked for stable, but should not be backported without this fix.
Richard Leitner [Mon, 6 Mar 2017 08:24:21 +0000 (09:24 +0100)]
usb: usb251xb: dt: add unit suffix to oc-delay and power-on-time
Rename oc-delay-* to oc-delay-us and make it expect a time value.
Furthermore add -ms suffix to power-on-time. There changes were
suggested by Rob Herring in https://lkml.org/lkml/2017/2/15/1283.
Remove the max_{power,current}_{sp,bp} properties of the usb251xb driver
from devicetree. This is done to simplify the dt bindings as requested
by Rob Herring in https://lkml.org/lkml/2017/2/15/1283. If those
properties are ever needed by somebody they can be enabled again easily.
Tobias Jakobi [Mon, 27 Feb 2017 23:46:58 +0000 (00:46 +0100)]
usb-storage: Add ignore-residue quirk for Initio INIC-3619
This USB-SATA bridge chip is used in a StarTech enclosure for
optical drives.
Without the quirk MakeMKV fails during the key exchange with an
installed BluRay drive:
> Error 'Scsi error - ILLEGAL REQUEST:COPY PROTECTION KEY EXCHANGE FAILURE - KEY NOT ESTABLISHED'
> occurred while issuing SCSI command AD010..080002400 to device 'SG:dev_11:2'
Johan Hovold [Tue, 7 Mar 2017 15:11:04 +0000 (16:11 +0100)]
USB: iowarrior: fix NULL-deref in write
Make sure to verify that we have the required interrupt-out endpoint for
IOWarrior56 devices to avoid dereferencing a NULL-pointer in write
should a malicious device lack such an endpoint.
Johan Hovold [Tue, 7 Mar 2017 15:11:03 +0000 (16:11 +0100)]
USB: iowarrior: fix NULL-deref at probe
Make sure to check for the required interrupt-in endpoint to avoid
dereferencing a NULL-pointer should a malicious device lack such an
endpoint.
Note that a fairly recent change purported to fix this issue, but added
an insufficient test on the number of endpoints only, a test which can
now be removed.
Fixes: 4ec0ef3a8212 ("USB: iowarrior: fix oops with malicious USB descriptors") Fixes: 946b960d13c1 ("USB: add driver for iowarrior devices.") Cc: stable <[email protected]> # 2.6.21 Signed-off-by: Johan Hovold <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>
The driver doesn't have a struct of_device_id table but supported devices
are registered via Device Trees. This is working on the assumption that a
I2C device registered via OF will always match a legacy I2C device ID and
that the MODALIAS reported will always be of the form i2c:<device>.
But this could change in the future so the correct approach is to have an
OF device ID table if the devices are registered via OF.
usb: ohci-at91: Do not drop unhandled USB suspend control requests
In patch 2e2aa1bc7eff90ecm, USB suspend and wakeup control requests are
passed to SFR_OHCIICR register. If a processor does not have such a
register, this hub control request will be dropped.
If no such a SFR register is available, all USB suspend control requests
will now be processed using ohci_hub_control()
(like before patch 2e2aa1bc7eff90ecm.)
Tested on an Atmel AT91SAM9G20 with an on-board TI TUSB2046B hub chip
If the last USB device is unplugged from the USB hub, the hub goes into
sleep and will not wakeup when an USB devices is inserted.
powerpc/powernv/ioda2: Update iommu table base on ownership change
On POWERNV platform, in order to do DMA via IOMMU (i.e. 32bit DMA in
our case), a device needs an iommu_table pointer set via
set_iommu_table_base().
The codeflow is:
- pnv_pci_ioda2_setup_dma_pe()
- pnv_pci_ioda2_setup_default_config()
- pnv_ioda_setup_bus_dma() [1]
pnv_pci_ioda2_setup_dma_pe() creates IOMMU groups,
pnv_pci_ioda2_setup_default_config() does default DMA setup,
pnv_ioda_setup_bus_dma() takes a bus PE (on IODA2, all physical function
PEs as bus PEs except NPU), walks through all underlying buses and
devices, adds all devices to an IOMMU group and sets iommu_table.
On IODA2, when VFIO is used, it takes ownership over a PE which means it
removes all tables and creates new ones (with a possibility of sharing
them among PEs). So when the ownership is returned from VFIO to
the kernel, the iommu_table pointer written to a device at [1] is
stale and needs an update.
This adds an "add_to_group" parameter to pnv_ioda_setup_bus_dma()
(in fact re-adds as it used to be there a while ago for different
reasons) to tell the helper if a device needs to be added to
an IOMMU group with an iommu_table update or just the latter.
This calls pnv_ioda_setup_bus_dma(..., false) from
pnv_ioda2_release_ownership() so when the ownership is restored,
32bit DMA can work again for a device. This does the same thing
on obtaining ownership as the iommu_table point is stale at this point
anyway and it is safer to have NULL there.
We did not hit this earlier as all tested devices in recent years were
only using 64bit DMA; the rare exception for this is MPT3 SAS adapter
which uses both 32bit and 64bit DMA access and it has not been tested
with VFIO much.
Linu Cherian [Wed, 8 Mar 2017 06:08:35 +0000 (11:38 +0530)]
KVM: arm64: Increase number of user memslots to 512
Having only 32 memslots is a real constraint for the maximum
number of PCI devices that can be assigned to a single guest.
Assuming each PCI device/virtual function having two memory BAR
regions, we could assign only 15 devices/virtual functions to a
guest.
Hence increase KVM_USER_MEM_SLOTS to 512 as done in other archs like
powerpc.
Merge tag 'fixes-for-v4.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb into usb-linus
Felipe writes:
usb: fixes for v4.11-rc2
dwc3 got a few fixes this time around:
Fixed an old bug where a broken endpoint descriptor passed in via
userspace through f_fs could prevent dwc3 from working because when
calculating max bursts, we could overwrite top 16 bits of a register.
Also fixed a bug on dwc3's ep_dequeue implementation which wasn't
properly incrementing our TRB dequeue pointer.
dwc3 on omap got two fixes: one for system suspend/resume and another
added a missing break statement on dwc3_omap_set_mailbox().
Apart from these, we have a set of smaller fixes including memory leak
in configfs, build warning fix in atmel udc and a revert of a broken
patch that went in during the merge window
powerpc/powernv/ioda2: Gracefully fail if too many TCE levels requested
The IODA2 specification says that a 64 DMA address cannot use top 4 bits
(3 are reserved and one is a "TVE select"); bottom page_shift bits
cannot be used for multilevel table addressing either.
The existing IODA2 table allocation code aligns the minimum TCE table
size to PAGE_SIZE so in the case of 64K system pages and 4K IOMMU pages,
we have 64-4-12=48 bits. Since 64K page stores 8192 TCEs, i.e. needs
13 bits, the maximum number of levels is 48/13 = 3 so we physically
cannot address more and EEH happens on DMA accesses.
This adds a check that too many levels were requested.
It is still possible to have 5 levels in the case of 4K system page size.
Which corresponds to the skb access after xmit handling. Fix by saving
skb->len and using the saved value to update stats.
Fixes: 193125dbd8eb2 ("net: Introduce VRF device driver") Signed-off-by: David Ahern <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Jarod Wilson [Mon, 6 Mar 2017 13:48:58 +0000 (08:48 -0500)]
team: use ETH_MAX_MTU as max mtu
This restores the ability to set a team device's mtu to anything higher
than 1500. Similar to the reported issue with bonding, the team driver
calls ether_setup(), which sets an initial max_mtu of 1500, while the
underlying hardware can handle something much larger. Just set it to
ETH_MAX_MTU to support all possible values, and the limitations of the
underlying devices will prevent setting anything too large.
net: ibm: emac: fix regression caused by emac_dt_phy_probe()
Julian Margetson reported a panic on his SAM460EX with Kernel 4.11-rc1:
| Unable to handle kernel paging request for data at address 0x00000014
| Oops: Kernel access of bad area, sig: 11 [#1]
| PREEMPT
| Canyonlands
| Modules linked in:
| CPU: 0 PID: 1 Comm: swapper Not tainted [...]
| task: ea838000 task.stack: ea836000
| NIP: c0599f5c LR: c0599dd8 CTR: 00000000
| REGS: ea837c80 TRAP: 0300 Not tainted [...]
| MSR: 00029000 <CE,EE,ME>
| CR: 24371242 XER: 20000000
| DEAR: 00000014 ESR: 00000000
| GPR00: c0599ce8ea837d30ea838000c0e52dccc0d56ffb [...]
| NIP [c0599f5c] emac_probe+0xfb4/0x1304
| LR [c0599dd8] emac_probe+0xe30/0x1304
| Call Trace:
| [ea837d30] [c0599ce8] emac_probe+0xd40/0x1304 (unreliable)
| [ea837d80] [c0533504] platform_drv_probe+0x48/0x90
| [ea837da0] [c0531c14] driver_probe_device+0x15c/0x2c4
| [ea837dd0] [c0531e04] __driver_attach+0x88/0xb0
| ---[ end trace ... ]---
The problem is caused by emac_dt_phy_probe() returing success (0)
for existing device-trees configurations that do not specify a
"phy-handle" property. This caused the code to skip the existing
phy probe and setup. Which led to essential phy related
data-structures being uninitialized.
This patch also removes the unused variable in emac_dt_phy_connect().
Fixes: a577ca6badb5261d ("net: emac: add support for device-tree based PHY discovery and setup") Reported-by: Julian Margetson <[email protected]> Signed-off-by: Christian Lamparter <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Cyril Bur [Tue, 7 Mar 2017 00:39:31 +0000 (11:39 +1100)]
selftests/powerpc: Replace stxvx and lxvx with stxvd2x/lxvd2x
On POWER8 (ISA 2.07) lxvx and stxvx are defined to be extended mnemonics
of lxvd2x and stxvd2x. For POWER9 (ISA 3.0) the HW architects in their
infinite wisdom made lxvx and stxvx instructions in their own right.
POWER9 aware GCC will use the POWER9 instruction for lxvx and stxvx
causing these selftests to fail on POWER8. Further compounding the
issue, because of the way -mvsx works it will cause the power9
instructions to be used regardless of -mcpu=power8 to GCC or -mpower8 to
AS.
The safest way to address the problem for now is to not use the extended
mnemonic. We don't care how the CPU loads the values from memory since
the tests only performs register comparisons, so using stdvd2x/lxvd2x
does not impact the test.
powerpc/perf: Fix perf_get_data_addr() for power9 DD1
Power9 DD1 do not support PMU_HAS_SIER flag and sdsync in
perf_get_data_addr() defaults to MMCRA_SDSYNC which is wrong. Since
power9 MMCRA does not support SDSYNC bit, patch includes PPMU_NO_SIAR
flag to the check and set the sdsync with MMCRA_SAMPLE_ENABLE;
Fixes: 27593d72c4ad ("powerpc/perf: Use MSR to report privilege level on P9 DD1") Signed-off-by: Madhavan Srinivasan <[email protected]> Signed-off-by: Michael Ellerman <[email protected]>
Tony Luck [Wed, 8 Mar 2017 17:35:39 +0000 (09:35 -0800)]
mm, page_alloc: Add missing check for memory holes
Commit 13ad59df67f1 ("mm, page_alloc: avoid page_to_pfn() when merging
buddies") moved the check for memory holes out of page_is_buddy() and
had the callers do the check.
But this wasn't done correctly in one place which caused ia64 to crash
very early in boot.
Update to fix that and make ia64 boot again.
[ v2: Vlastimil pointed out we don't need to call page_to_pfn()
since we already have the result of that in "buddy_pfn" ]
Linus Torvalds [Wed, 8 Mar 2017 19:06:05 +0000 (11:06 -0800)]
Merge tag 'ktest-v4.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-ktest
Pull ktest fixes from Steven Rostedt:
"Greg Kroah-Hartman reported to me that the ktest of v4.11-rc1 locked
up in an infinite loop while doing the make mrproper.
Looking into the cause I noticed that a recent update to the function
run_command (used for running all shell commands, including "make
mrproper") changed the internal loop to use the function
wait_for_input.
The wait_for_input function uses select to look at two file
descriptors. One is the file descriptor of the command it is running,
the other is STDIN. The STDIN check was not checking the return status
of the sysread call, and was also just writing a lot of data into
syswrite without regard to the size of the data read.
Changing the code to check the return status of sysread, and also to
still process the passed in descriptor data without looping back to
the select fixed Greg's problem.
While looking at this code I also realized that the loop did not honor
the timeout if STDIN always had input (or for some reason return
error). this could prevent wait_for_input to timeout on the file
descriptor it is suppose to be waiting for. That is fixed too"
* tag 'ktest-v4.11-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-ktest:
ktest: Make sure wait_for_input does honor the timeout
ktest: Fix while loop in wait_for_input
Linus Torvalds [Wed, 8 Mar 2017 18:42:13 +0000 (10:42 -0800)]
overlayfs: remove now unnecessary header file include
This removes the extra include header file that was added in commit e58bc927835a "Pull overlayfs updates from Miklos Szeredi" now that it
is no longer needed.
There are probably other such includes that got added during the
scheduler header splitup series, but this is the one that annoyed me
personally and I know about.
xfs: try any AG when allocating the first btree block when reflinking
When a reflink operation causes the bmap code to allocate a btree block
we're currently doing single-AG allocations due to having ->firstblock
set and then try any higher AG due a little reflink quirk we've put in
when adding the reflink code. But given that we do not have a minleft
reservation of any kind in this AG we can still not have any space in
the same or higher AG even if the file system has enough free space.
To fix this use a XFS_ALLOCTYPE_FIRST_AG allocation in this fall back
path instead.
[And yes, we need to redo this properly instead of piling hacks over
hacks. I'm working on that, but it's not going to be a small series.
In the meantime this fixes the customer reported issue]
Also add a warning for failing allocations to make it easier to debug.
Linus Torvalds [Tue, 7 Mar 2017 23:33:14 +0000 (15:33 -0800)]
sched/headers: fix up header file dependency on <linux/sched/signal.h>
The scheduler header file split and cleanups ended up exposing a few
nasty header file dependencies, and in particular it showed how we in
<linux/wait.h> ended up depending on "signal_pending()", which now comes
from <linux/sched/signal.h>.
That's a very subtle and annoying dependency, which already caused a
semantic merge conflict (see commit e58bc927835a "Pull overlayfs updates
from Miklos Szeredi", which added that fixup in the merge commit).
It turns out that we can avoid this dependency _and_ improve code
generation by moving the guts of the fairly nasty helper #define
__wait_event_interruptible_locked() to out-of-line code. The code that
includes the signal_pending() check is all in the slow-path where we
actually go to sleep waiting for the event anyway, so using a helper
function is the right thing to do.
Using a helper function is also what we already did for the non-locked
versions, see the "__wait_event*()" macros and the "prepare_to_wait*()"
set of helper functions.
We might want to try to unify all these macro games, we have a _lot_ of
subtly different wait-event loops. But this is the minimal patch to fix
the annoying header dependency.
Brian Foster [Wed, 8 Mar 2017 17:58:08 +0000 (09:58 -0800)]
xfs: use iomap new flag for newly allocated delalloc blocks
Commit fa7f138 ("xfs: clear delalloc and cache on buffered write
failure") fixed one regression in the iomap error handling code and
exposed another. The fundamental problem is that if a buffered write
is a rewrite of preexisting delalloc blocks and the write fails, the
failure handling code can punch out preexisting blocks with valid
file data.
This was reproduced directly by sub-block writes in the LTP
kernel/syscalls/write/write03 test. A first 100 byte write allocates
a single block in a file. A subsequent 100 byte write fails and
punches out the block, including the data successfully written by
the previous write.
To address this problem, update the ->iomap_begin() handler to
distinguish newly allocated delalloc blocks from preexisting
delalloc blocks via the IOMAP_F_NEW flag. Use this flag in the
->iomap_end() handler to decide when a failed or short write should
punch out delalloc blocks.
This introduces the subtle requirement that ->iomap_begin() should
never combine newly allocated delalloc blocks with existing blocks
in the resulting iomap descriptor. This can occur when a new
delalloc reservation merges with a neighboring extent that is part
of the current write, for example. Therefore, drop the
post-allocation extent lookup from xfs_bmapi_reserve_delalloc() and
just return the record inserted into the fork. This ensures only new
blocks are returned and thus that preexisting delalloc blocks are
always handled as "found" blocks and not punched out on a failed
rewrite.
NeilBrown [Tue, 7 Mar 2017 20:38:05 +0000 (07:38 +1100)]
blk: improve order of bio handling in generic_make_request()
To avoid recursion on the kernel stack when stacked block devices
are in use, generic_make_request() will, when called recursively,
queue new requests for later handling. They will be handled when the
make_request_fn for the current bio completes.
If any bios are submitted by a make_request_fn, these will ultimately
be handled seqeuntially. If the handling of one of those generates
further requests, they will be added to the end of the queue.
This strict first-in-first-out behaviour can lead to deadlocks in
various ways, normally because a request might need to wait for a
previous request to the same device to complete. This can happen when
they share a mempool, and can happen due to interdependencies
particular to the device. Both md and dm have examples where this happens.
These deadlocks can be erradicated by more selective ordering of bios.
Specifically by handling them in depth-first order. That is: when the
handling of one bio generates one or more further bios, they are
handled immediately after the parent, before any siblings of the
parent. That way, when generic_make_request() calls make_request_fn
for some particular device, we can be certain that all previously
submited requests for that device have been completely handled and are
not waiting for anything in the queue of requests maintained in
generic_make_request().
An easy way to achieve this would be to use a last-in-first-out stack
instead of a queue. However this will change the order of consecutive
bios submitted by a make_request_fn, which could have unexpected consequences.
Instead we take a slightly more complex approach.
A fresh queue is created for each call to a make_request_fn. After it completes,
any bios for a different device are placed on the front of the main queue, followed
by any bios for the same device, followed by all bios that were already on
the queue before the make_request_fn was called.
This provides the depth-first approach without reordering bios on the same level.
This, by itself, it not enough to remove all deadlocks. It just makes
it possible for drivers to take the extra step required themselves.
To avoid deadlocks, drivers must never risk waiting for a request
after submitting one to generic_make_request. This includes never
allocing from a mempool twice in the one call to a make_request_fn.
A common pattern in drivers is to call bio_split() in a loop, handling
the first part and then looping around to possibly split the next part.
Instead, a driver that finds it needs to split a bio should queue
(with generic_make_request) the second part, handle the first part,
and then return. The new code in generic_make_request will ensure the
requests to underlying bios are processed first, then the second bio
that was split off. If it splits again, the same process happens. In
each case one bio will be completely handled before the next one is attempted.
With this is place, it should be possible to disable the
punt_bios_to_recover() recovery thread for many block devices, and
eventually it may be possible to remove it completely.
Jan Kara [Wed, 8 Mar 2017 16:48:34 +0000 (17:48 +0100)]
Revert "scsi, block: fix duplicate bdi name registration crashes"
This reverts commit 0dba1314d4f81115dce711292ec7981d17231064. It causes
leaking of device numbers for SCSI when SCSI registers multiple gendisks
for one request_queue in succession. It can be easily reproduced using
Omar's script [1] on kernel with CONFIG_DEBUG_TEST_DRIVER_REMOVE.
Furthermore the protection provided by this commit is not needed anymore
as the problem it was fixing got also fixed by commit 165a5e22fafb
"block: Move bdi_unregister() to del_gendisk()".
Jan Kara [Wed, 8 Mar 2017 16:48:33 +0000 (17:48 +0100)]
block: Make del_gendisk() safer for disks without queues
Commit 165a5e22fafb "block: Move bdi_unregister() to del_gendisk()"
added disk->queue dereference to del_gendisk(). Although del_gendisk()
is not supposed to be called without disk->queue valid and
blk_unregister_queue() warns in that case, this change will make it oops
instead. Return to the old more robust behavior of just warning when
del_gendisk() gets called for gendisk with disk->queue being NULL.
Jan Kara [Wed, 8 Mar 2017 16:48:32 +0000 (17:48 +0100)]
bdi: Fix use-after-free in wb_congested_put()
bdi_writeback_congested structures get created for each blkcg and bdi
regardless whether bdi is registered or not. When they are created in
unregistered bdi and the request queue (and thus bdi) is then destroyed
while blkg still holds reference to bdi_writeback_congested structure,
this structure will be referencing freed bdi and last wb_congested_put()
will try to remove the structure from already freed bdi.
With commit 165a5e22fafb "block: Move bdi_unregister() to
del_gendisk()", SCSI started to destroy bdis without calling
bdi_unregister() first (previously it was calling bdi_unregister() even
for unregistered bdis) and thus the code detaching
bdi_writeback_congested in cgwb_bdi_destroy() was not triggered and we
started hitting this use-after-free bug. It is enough to boot a KVM
instance with virtio-scsi device to trigger this behavior.
Fix the problem by detaching bdi_writeback_congested structures in
bdi_exit() instead of bdi_unregister(). This is also more logical as
they can get attached to bdi regardless whether it ever got registered
or not.
projects / linux.git / log