When PTP timestamping is enabled on Tx, the controller
inserts the Tx timestamp at the beginning of the frame
buffer, between SFD and the L2 frame header. This means
that the skb provided by the stack is required to have
enough headroom otherwise a new skb needs to be created
by the driver to accommodate the timestamp inserted by h/w.
Up until now the driver was relying on the second option,
using skb_realloc_headroom() to create a new skb to accommodate
PTP frames. Turns out that this method is not reliable, as
reallocation of skbs for PTP frames along with the required
overhead (skb_set_owner_w, consume_skb) is causing random
crashes in subsequent skb_*() calls, when multiple concurrent
TCP streams are run at the same time on the same device
(as seen in James' report).
Note that these crashes don't occur with a single TCP stream,
nor with multiple concurrent UDP streams, but only when multiple
TCP streams are run concurrently with the PTP packet flow
(doing skb reallocation).
This patch enforces the first method, by requesting enough
headroom from the stack to accommodate PTP frames, and so avoiding
skb_realloc_headroom() & co, and the crashes no longer occur.
There's no reason not to set needed_headroom to a large enough
value to accommodate PTP frames, so in this regard this patch
is a fix.
Reported-by: James Jurack <james.jurack@ametek.com> Fixes: bee9e58c9e98 ("gianfar:don't add FCB length to hard_header_len") Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com> Link: https://lore.kernel.org/r/20201020173605.1173-1-claudiu.manoil@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When PTP timestamping is enabled on Tx, the controller
inserts the Tx timestamp at the beginning of the frame
buffer, between SFD and the L2 frame header. This means
that the skb provided by the stack is required to have
enough headroom otherwise a new skb needs to be created
by the driver to accommodate the timestamp inserted by h/w.
Up until now the driver was relying on skb_realloc_headroom()
to create new skbs to accommodate PTP frames. Turns out that
this method is not reliable in this context at least, as
skb_realloc_headroom() for PTP frames can cause random crashes,
mostly in subsequent skb_*() calls, when multiple concurrent
TCP streams are run at the same time with the PTP flow
on the same device (as seen in James' report). I also noticed
that when the system is loaded by sending multiple TCP streams,
the driver receives cloned skbs in large numbers.
skb_cow_head() instead proves to be stable in this scenario,
and not only handles cloned skbs too but it's also more efficient
and widely used in other drivers.
The commit introducing skb_realloc_headroom in the driver
goes back to 2009, commit 93c1285c5d92
("gianfar: reallocate skb when headroom is not enough for fcb").
For practical purposes I'm referencing a newer commit (from 2012)
that brings the code to its current structure (and fixes the PTP
case).
Fixes: 9c4886e5e63b ("gianfar: Fix invalid TX frames returned on error queue when time stamping") Reported-by: James Jurack <james.jurack@ametek.com> Suggested-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com> Link: https://lore.kernel.org/r/20201029081057.8506-1-claudiu.manoil@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Correct skb refcount in alloc_ctrl_skb(), causing skb memleak
when chtls_send_abort() called with NULL skb.
it was always leaking the skb, correct it by incrementing skb
refs by one.
work request skb queued in chtls_setkey() won't be freed
because resources are already cleaned for this connection,
fix it by not queuing work request while socket is closing.
v1->v2:
- fix W=1 warning.
v2->v3:
- separate it out from another memleak fix.
In my test setup, I had a SAMA5D27 device configured with ip forwarding, and
second device with usb ethernet (r8152) sending ICMP packets. If the packet
was larger than about 220 bytes, the SAMA5 device would "oops" with the
following trace:
kernel BUG at net/core/skbuff.c:1863!
Internal error: Oops - BUG: 0 [#1] ARM
Modules linked in: xt_MASQUERADE ppp_async ppp_generic slhc iptable_nat xt_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 can_raw can bridge stp llc ipt_REJECT nf_reject_ipv4 sd_mod cdc_ether usbnet usb_storage r8152 scsi_mod mii o
ption usb_wwan usbserial micrel macb at91_sama5d2_adc phylink gpio_sama5d2_piobu m_can_platform m_can industrialio_triggered_buffer kfifo_buf of_mdio can_dev fixed_phy sdhci_of_at91 sdhci_pltfm libphy sdhci mmc_core ohci_at91 ehci_atmel o
hci_hcd iio_rescale industrialio sch_fq_codel spidev prox2_hal(O)
CPU: 0 PID: 0 Comm: swapper Tainted: G O 5.9.1-prox2+ #1
Hardware name: Atmel SAMA5
PC is at skb_put+0x3c/0x50
LR is at macb_start_xmit+0x134/0xad0 [macb]
pc : [<c05258cc>] lr : [<bf0ea5b8>] psr: 20070113
sp : c0d01a60 ip : c07232c0 fp : c4250000
r10: c0d03cc8 r9 : 00000000 r8 : c0d038c0
r7 : 00000000 r6 : 00000008 r5 : c59b66c0 r4 : 0000002a
r3 : 8f659eff r2 : c59e9eea r1 : 00000001 r0 : c59b66c0
Flags: nzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none
Control: 10c53c7d Table: 2640c059 DAC: 00000051
Process swapper (pid: 0, stack limit = 0x75002d81)
The solution was to force nonlinear buffers to be cloned. This was previously
reported by Klaus Doth (https://www.spinics.net/lists/netdev/msg556937.html)
but never formally submitted as a patch.
This is the third revision, hopefully the formatting is correct this time!
Suggested-by: Klaus Doth <krnl@doth.eu> Fixes: 653e92a9175e ("net: macb: add support for padding and fcs computation") Signed-off-by: Mark Deneen <mdeneen@saucontech.com> Link: https://lore.kernel.org/r/20201030155814.622831-1-mdeneen@saucontech.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
int status;
int thread = waitpid(-1, &status, 0);
assert(thread > 0 && thread != pid);
assert(status == 0x80137f);
return 0;
}
fails and triggers WARN_ON_ONCE(!signr) in do_jobctl_trap().
This is because task_join_group_stop() has 2 problems when current is traced:
1. We can't rely on the "JOBCTL_STOP_PENDING" check, a stopped tracee
can be woken up by debugger and it can clone another thread which
should join the group-stop.
We need to check group_stop_count || SIGNAL_STOP_STOPPED.
2. If SIGNAL_STOP_STOPPED is already set, we should not increment
sig->group_stop_count and add JOBCTL_STOP_CONSUME. The new thread
should stop without another do_notify_parent_cldstop() report.
To clarify, the problem is very old and we should blame
ptrace_init_task(). But now that we have task_join_group_stop() it makes
more sense to fix this helper to avoid the code duplication.
==================================================================
BUG: KASAN: use-after-free in tipc_bcast_get_mode+0x3ab/0x400 net/tipc/bcast.c:759
Read of size 1 at addr ffff88805e6b3571 by task kworker/0:6/3850
CPU: 0 PID: 3850 Comm: kworker/0:6 Not tainted 5.8.0-rc7-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events tipc_net_finalize_work
Fix it by calling flush_scheduled_work() to make sure the
tipc_net_finalize_work() stopped before releasing bcbase object.
Reported-by: syzbot+6ea1f7a8df64596ef4d7@syzkaller.appspotmail.com Reported-by: syzbot+e9cc557752ab126c1b99@syzkaller.appspotmail.com Acked-by: Jon Maloy <jmaloy@redhat.com> Signed-off-by: Hoang Huu Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 39d114ddc682 ("arm64: add KASAN support") added .weak directives to
arch/arm64/lib/mem*.S instead of changing the existing SYM_FUNC_START_PI
macros. This can lead to the assembly snippet `.weak memcpy ... .globl
memcpy` which will produce a STB_WEAK memcpy with GNU as but STB_GLOBAL
memcpy with LLVM's integrated assembler before LLVM 12. LLVM 12 (since
https://reviews.llvm.org/D90108) will error on such an overridden symbol
binding.
Use the appropriate SYM_FUNC_START_WEAK_PI instead.
Fixes: 39d114ddc682 ("arm64: add KASAN support") Reported-by: Sami Tolvanen <samitolvanen@google.com> Signed-off-by: Fangrui Song <maskray@google.com> Tested-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Cc: <stable@vger.kernel.org> Link: https://lore.kernel.org/r/20201029181951.1866093-1-maskray@google.com Signed-off-by: Will Deacon <will@kernel.org> Cc: Jian Cai <jiancai@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In an effort to clarify and simplify the annotation of assembly functions
in the kernel new macros have been introduced. These replace ENTRY and
ENDPROC and also add a new annotation for static functions which previously
had no ENTRY equivalent. Update the annotations in the library code to the
new macros.
Signed-off-by: Mark Brown <broonie@kernel.org>
[will: Use SYM_FUNC_START_WEAK_PI] Signed-off-by: Will Deacon <will@kernel.org> Cc: Jian Cai <jiancai@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As part of an effort to make the annotations in assembly code clearer and
more consistent new macros have been introduced, including replacements
for ENTRY() and ENDPROC().
On arm64 we have ENDPIPROC(), a custom version of ENDPROC() which is
used for code that will need to run in position independent environments
like EFI, it creates an alias for the function with the prefix __pi_ and
then emits the standard ENDPROC. Add new-style macros to replace this
which expand to the standard SYM_FUNC_*() and SYM_FUNC_ALIAS_*(),
resulting in the same object code. These are added in linkage.h for
consistency with where the generic assembler code has its macros.
Signed-off-by: Mark Brown <broonie@kernel.org>
[will: Rename 'WEAK' macro, use ';' instead of ASM_NL, deprecate ENDPIPROC] Signed-off-by: Will Deacon <will@kernel.org> Cc: Jian Cai <jiancai@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Introduce new C macros for annotations of functions and data in
assembly. There is a long-standing mess in macros like ENTRY, END,
ENDPROC and similar. They are used in different manners and sometimes
incorrectly.
So introduce macros with clear use to annotate assembly as follows:
a) Support macros for the ones below
SYM_T_FUNC -- type used by assembler to mark functions
SYM_T_OBJECT -- type used by assembler to mark data
SYM_T_NONE -- type used by assembler to mark entries of unknown type
They are defined as STT_FUNC, STT_OBJECT, and STT_NOTYPE
respectively. According to the gas manual, this is the most portable
way. I am not sure about other assemblers, so this can be switched
back to %function and %object if this turns into a problem.
Architectures can also override them by something like ", @function"
if they need.
SYM_A_ALIGN, SYM_A_NONE -- align the symbol?
SYM_L_GLOBAL, SYM_L_WEAK, SYM_L_LOCAL -- linkage of symbols
b) Mostly internal annotations, used by the ones below
SYM_ENTRY -- use only if you have to (for non-paired symbols)
SYM_START -- use only if you have to (for paired symbols)
SYM_END -- use only if you have to (for paired symbols)
c) Annotations for code
SYM_INNER_LABEL_ALIGN -- only for labels in the middle of code
SYM_INNER_LABEL -- only for labels in the middle of code
SYM_FUNC_START_LOCAL_ALIAS -- use where there are two local names for
one function
SYM_FUNC_START_ALIAS -- use where there are two global names for one
function
SYM_FUNC_END_ALIAS -- the end of LOCAL_ALIASed or ALIASed function
SYM_FUNC_START -- use for global functions
SYM_FUNC_START_NOALIGN -- use for global functions, w/o alignment
SYM_FUNC_START_LOCAL -- use for local functions
SYM_FUNC_START_LOCAL_NOALIGN -- use for local functions, w/o
alignment
SYM_FUNC_START_WEAK -- use for weak functions
SYM_FUNC_START_WEAK_NOALIGN -- use for weak functions, w/o alignment
SYM_FUNC_END -- the end of SYM_FUNC_START_LOCAL, SYM_FUNC_START,
SYM_FUNC_START_WEAK, ...
For functions with special (non-C) calling conventions:
SYM_CODE_START -- use for non-C (special) functions
SYM_CODE_START_NOALIGN -- use for non-C (special) functions, w/o
alignment
SYM_CODE_START_LOCAL -- use for local non-C (special) functions
SYM_CODE_START_LOCAL_NOALIGN -- use for local non-C (special)
functions, w/o alignment
SYM_CODE_END -- the end of SYM_CODE_START_LOCAL or SYM_CODE_START
d) For data
SYM_DATA_START -- global data symbol
SYM_DATA_START_LOCAL -- local data symbol
SYM_DATA_END -- the end of the SYM_DATA_START symbol
SYM_DATA_END_LABEL -- the labeled end of SYM_DATA_START symbol
SYM_DATA -- start+end wrapper around simple global data
SYM_DATA_LOCAL -- start+end wrapper around simple local data
==========
The macros allow to pair starts and ends of functions and mark functions
correctly in the output ELF objects.
All users of the old macros in x86 are converted to use these in further
patches.
Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Len Brown <len.brown@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: linux-arch@vger.kernel.org Cc: linux-doc@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: linux-pm@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will@kernel.org> Cc: x86-ml <x86@kernel.org> Cc: xen-devel@lists.xenproject.org Link: https://lkml.kernel.org/r/20191011115108.12392-2-jslaby@suse.cz Cc: Jian Cai <jiancai@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add alternative topology binary file name based on used machine driver
and fallback to use this name after failed attempt to load topology file
with name based on NHLT.
This change addresses multiple issues with current mechanism, for
example - there are devices without NHLT table, and that currently
results in tplg_name being empty.
The "mmio" writes into vgpu registers are simple memory traps from the
guest into the host. We do not need to assert in the guest that the
device is awake for the io as we do not write to the device itself.
However, over time we have refactored all the mmio accessors with the
result that the vgpu reuses the gen2 accessors and so inherits the
assert for runtime-pm of the native device. The assert though has
actually been there since commit 3be0bf5acca6 ("drm/i915: Create vGPU
specific MMIO operations to reduce traps").
References: 3be0bf5acca6 ("drm/i915: Create vGPU specific MMIO operations to reduce traps") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Yan Zhao <yan.y.zhao@intel.com> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com> Cc: stable@vger.kernel.org Link: https://patchwork.freedesktop.org/patch/msgid/20200811092532.13753-1-chris@chris-wilson.co.uk
(cherry picked from commit 0e65ce24a33c1d37da4bf43c34e080334ec6cb60) Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When running gem_exec_nop, it floods the system with many requests (with
the goal of userspace submitting faster than the HW can process a single
empty batch). This causes the driver to continually resubmit new
requests onto the end of an active context, a flood of lite-restore
preemptions. If we time this just right, Tigerlake hangs.
Inserting a small delay between the processing of CS events and
submitting the next context, prevents the hang. Naturally it does not
occur with debugging enabled. The suspicion then is that this is related
to the issues with the CS event buffer, and inserting an mmio read of
the CS pointer status appears to be very successful in preventing the
hang. Other registers, or uncached reads, or plain mb, do not prevent
the hang, suggesting that register is key -- but that the hang can be
prevented by a simple udelay, suggests it is just a timing issue like
that encountered by commit 233c1ae3c83f ("drm/i915/gt: Wait for CSB
entries on Tigerlake"). Also note that the hang is not prevented by
applying CTX_DESC_FORCE_RESTORE, or by inserting a delay on the GPU
between requests.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Bruce Chang <yu.bruce.chang@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: stable@vger.kernel.org Acked-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20201015195023.32346-1-chris@chris-wilson.co.uk
(cherry picked from commit 6ca7217dffaf1abba91558e67a2efb655ac91405) Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As the error capture will compress user buffers as directed to by the
user, it can take an arbitrary amount of time and space. Break up the
compression loops with a call to cond_resched(), that will allow other
processes to schedule (avoiding the soft lockups) and also serve as a
warning should we try to make this loop atomic in the future.
Currently in case of alignment or FCS error if the packet cannot be
corrected it's still not dropped. Report the error properly and drop the
packet while making the code around a little bit more readable.
The "cb_pcidas" driver supports asynchronous commands on the analog
output (AO) subdevice for those boards that have an AO FIFO. The code
(in `cb_pcidas_ao_check_chanlist()` and `cb_pcidas_ao_cmd()`) to
validate and set up the command supports output to a single channel or
to two channels simultaneously (the boards have two AO channels).
However, the code in `cb_pcidas_auto_attach()` that initializes the
subdevices neglects to initialize the AO subdevice's `len_chanlist`
member, leaving it set to 0, but the Comedi core will "correct" it to 1
if the driver neglected to set it. This limits commands to use a single
channel (either channel 0 or 1), but the limit should be two channels.
Set the AO subdevice's `len_chanlist` member to be the same value as the
`n_chan` member, which will be 2.
The DBGD{CCINT,SCRext} and DBGVCR register entries in the cp14 array
are missing their target register, resulting in all accesses being
targetted at the guard sysreg (indexed by __INVALID_SYSREG__).
Point the emulation code at the actual register entries.
It appears that firmware nodes can be shared between devices. In such case
when a (child) device is about to be deleted, its firmware node may be shared
and ACPI_COMPANION_SET(..., NULL) call for it breaks the secondary link
of the shared primary firmware node.
In order to prevent that, check, if the device has a parent and parent's
firmware node is shared with its child, and avoid crashing the link.
Fixes: c15e1bdda436 ("device property: Fix the secondary firmware node handling in set_primary_fwnode()") Reported-by: Ferry Toth <fntoth@gmail.com> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Tested-by: Ferry Toth <fntoth@gmail.com> Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Behind primary and secondary we understand the type of the nodes
which might define their ordering. However, if primary node gone,
we can't maintain the ordering by definition of the linked list.
Thus, by ordering secondary node becomes first in the list.
But in this case the meaning of it is still secondary (or auxiliary).
The type of the node is maintained by the secondary pointer in it:
secondary pointer Meaning
NULL or valid primary node
ERR_PTR(-ENODEV) secondary node
So, if by some reason we do the following sequence of calls
This concept is supported by the description of set_primary_fwnode()
along with implementation of set_secondary_fwnode(). Hence, fix
the commit c15e1bdda436 to follow this as well.
Fixes: c15e1bdda436 ("device property: Fix the secondary firmware node handling in set_primary_fwnode()") Cc: Ferry Toth <fntoth@gmail.com> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Tested-by: Ferry Toth <fntoth@gmail.com> Cc: 5.9+ <stable@vger.kernel.org> # 5.9+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit f6361c6b3880 ("ARM: S3C24XX: remove separate restart code")
removed usage of the watchdog reset platform code in favor of the
Samsung SoC watchdog driver. However the latter was not selected thus
S3C24xx platforms lost reset abilities.
Cc: <stable@vger.kernel.org> Fixes: f6361c6b3880 ("ARM: S3C24XX: remove separate restart code") Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Selecting CONFIG_SAMSUNG_PM_DEBUG (depending on CONFIG_DEBUG_LL) but
without CONFIG_MMU leads to build errors:
arch/arm/plat-samsung/pm-debug.c: In function ‘s3c_pm_uart_base’:
arch/arm/plat-samsung/pm-debug.c:57:2: error:
implicit declaration of function ‘debug_ll_addr’ [-Werror=implicit-function-declaration]
Fixes: 99b2fc2b8b40 ("ARM: SAMSUNG: Use debug_ll_addr() to get UART base address") Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Cc: <stable@vger.kernel.org> Link: https://lore.kernel.org/r/20200910154150.3318-1-krzk@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When starting a HP machine with HIL driver but without an HIL keyboard
or HIL mouse attached, it may happen that data written to the HIL loop
gets stuck (e.g. because the transaction queue is full). Usually one
will then have to reboot the machine because all you see is and endless
output of:
Transaction add failed: transaction already queued?
In the higher layers hp_sdc_enqueue_transaction() is called to queued up
a HIL packet. This function returns an error code, and this patch adds
the necessary checks for this return code and disables the HIL driver if
further packets can't be sent.
Some of the font tty ioctl's always used the current foreground VC for
their operations. Don't do that then.
This fixes a data race on fg_console.
Side note: both Michael Ellerman and Jiri Slaby point out that all these
ioctls are deprecated, and should probably have been removed long ago,
and everything seems to be using the KDFONTOP ioctl instead.
In fact, Michael points out that it looks like busybox's loadfont
program seems to have switched over to using KDFONTOP exactly _because_
of this bug (ahem.. 12 years ago ;-).
sdhci-of-dwcmshc meets an eMMC read performance regression with below
command after commit 427b6514d095 ("mmc: sdhci: Add Auto CMD Auto
Select support"):
Before the commit, the above command gives 120MB/s
After the commit, the above command gives 51.3 MB/s
So it looks like sdhci-of-dwcmshc expects Version 4 Mode for Auto
CMD Auto Select. Fix the performance degradation by ensuring v4_mode
is true to use Auto CMD Auto Select.
Fixes: 427b6514d095 ("mmc: sdhci: Add Auto CMD Auto Select support") Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20201015174115.4cf2c19a@xhacker.debian Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mmc0: Tuning timeout, falling back to fixed sampling clock
There are SD cards which takes a significant longer time to reply to the
first CMD19 command. The eSDHC takes the data timeout value into account
during the tuning period. The SDHCI core doesn't explicitly set this
timeout for the tuning procedure. Thus on the slow cards, there might be
a spurious "Buffer Read Ready" interrupt, which in turn triggers a wrong
sequence of events. In the end this will lead to an unsuccessful tuning
procedure and to the above error.
To workaround this, set the timeout to the maximum value (which is the
best we can do) and the SDHCI core will take care of the proper timeout
handling.
During the stability test, there are some errors:
ext4_lookup:1590: inode #6967: comm fsstress: iget: checksum invalid.
If the inode->i_iblocks too big and doesn't set huge file flag, checksum
will not be recalculated when update the inode information to it's buffer.
If other inode marks the buffer dirty, then the inconsistent inode will
be flushed to disk.
Fix this problem by checking i_blocks in advance.
Cc: stable@kernel.org Signed-off-by: Luo Meng <luomeng12@huawei.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Link: https://lore.kernel.org/r/20201020013631.3796673-1-luomeng12@huawei.com Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If riov and wiov are both defined and they point to different
objects, only riov is initialized. If the wiov is not initialized
by the caller, the function fails returning -EINVAL and printing
"Readable desc 0x... after writable" error message.
This issue happens when descriptors have both readable and writable
buffers (eg. virtio-blk devices has virtio_blk_outhdr in the readable
buffer and status as last byte of writable buffer) and we call
__vringh_iov() to get both type of buffers in two different iovecs.
Let's replace the 'else if' clause with 'if' to initialize both
riov and wiov if they are not NULL.
As checkpatch pointed out, we also avoid crashing the kernel
when riov and wiov are both NULL, replacing BUG() with WARN_ON()
and returning -EINVAL.
Fixes: f87d0fbb5798 ("vringh: host-side implementation of virtio rings.") Cc: stable@vger.kernel.org Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> Link: https://lore.kernel.org/r/20201008204256.162292-1-sgarzare@redhat.com Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We don't need to check the new buffer size, and the return value
had confused resize_buffer_duplicate_size().
...
ret = ring_buffer_resize(trace_buf->buffer,
per_cpu_ptr(size_buf->data,cpu_id)->entries, cpu_id);
if (ret == 0)
per_cpu_ptr(trace_buf->data, cpu_id)->entries =
per_cpu_ptr(size_buf->data, cpu_id)->entries;
...
Link: https://lkml.kernel.org/r/20201019142242.11560-1-hqjagain@gmail.com Cc: stable@vger.kernel.org Fixes: d60da506cbeb3 ("tracing: Add a resize function to make one buffer equivalent to another buffer") Signed-off-by: Qiujun Huang <hqjagain@gmail.com> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
con->out_msg must be cleared on Policy::stateful_server
(!CEPH_MSG_CONNECT_LOSSY) faults. Not doing so botches the
reconnection attempt, because after writing the banner the
messenger moves on to writing the data section of that message
(either from where it got interrupted by the connection reset or
from the beginning) instead of writing struct ceph_msg_connect.
This results in a bizarre error message because the server
sends CEPH_MSGR_TAG_BADPROTOVER but we think we wrote struct
ceph_msg_connect:
libceph: mds0 (1)172.21.15.45:6828 socket error on write
ceph: mds0 reconnect start
libceph: mds0 (1)172.21.15.45:6829 socket closed (con state OPEN)
libceph: mds0 (1)172.21.15.45:6829 protocol version mismatch, my 32 != server's 32
libceph: mds0 (1)172.21.15.45:6829 protocol version mismatch
AFAICT this bug goes back to the dawn of the kernel client.
The reason it survived for so long is that only MDS sessions
are stateful and only two MDS messages have a data section:
CEPH_MSG_CLIENT_RECONNECT (always, but reconnecting is rare)
and CEPH_MSG_CLIENT_REQUEST (only when xattrs are involved).
The connection has to get reset precisely when such message
is being sent -- in this case it was the former.
Currently both error code paths handled in dal_gpio_open_ex() issues
ASSERT_CRITICAL(), and this leads to a kernel panic unnecessarily if
CONFIG_KGDB is enabled. Since basically both are non-critical errors
and can be recovered, drop those assert calls and use a safer one,
BREAK_TO_DEBUGGER(), for allowing the debugging, instead.
BugLink: https://bugzilla.opensuse.org/show_bug.cgi?id=1177973 Cc: <stable@vger.kernel.org> Acked-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ASSERT_CRITICAL() invokes kgdb_breakpoint() whenever either
CONFIG_KGDB or CONFIG_HAVE_KGDB is set. This, however, may lead to a
kernel panic when no kdb stuff is attached, since the
kgdb_breakpoint() call issues INT3. It's nothing but a surprise for
normal end-users.
For avoiding the pitfall, make the kgdb_breakpoint() call only when
CONFIG_DEBUG_KERNEL_DC is set.
https://bugzilla.opensuse.org/show_bug.cgi?id=1177973 Cc: <stable@vger.kernel.org> Acked-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On connector destruction call drm_dp_mst_topology_mgr_destroy
to release resources allocated in drm_dp_mst_topology_mgr_init.
Do it only if MST manager was initilized before otherwsie a crash
is seen on driver unload/device unplug.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Suggested-by: Christian König <christian.koenig@amd.com> Signed-off-by: Madhav Chauhan <madhav.chauhan@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If interrupt comes late, during probe error path or device remove (could
be triggered with CONFIG_DEBUG_SHIRQ), the interrupt handler
i2c_imx_isr() will access registers with the clock being disabled. This
leads to external abort on non-linefetch on Toradex Colibri VF50 module
(with Vybrid VF5xx):
Unhandled fault: external abort on non-linefetch (0x1008) at 0x8882d003
Internal error: : 1008 [#1] ARM
Modules linked in:
CPU: 0 PID: 1 Comm: swapper Not tainted 5.7.0 #607
Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
(i2c_imx_isr) from [<8017009c>] (free_irq+0x25c/0x3b0)
(free_irq) from [<805844ec>] (release_nodes+0x178/0x284)
(release_nodes) from [<80580030>] (really_probe+0x10c/0x348)
(really_probe) from [<80580380>] (driver_probe_device+0x60/0x170)
(driver_probe_device) from [<80580630>] (device_driver_attach+0x58/0x60)
(device_driver_attach) from [<805806bc>] (__driver_attach+0x84/0xc0)
(__driver_attach) from [<8057e228>] (bus_for_each_dev+0x68/0xb4)
(bus_for_each_dev) from [<8057f3ec>] (bus_add_driver+0x144/0x1ec)
(bus_add_driver) from [<80581320>] (driver_register+0x78/0x110)
(driver_register) from [<8010213c>] (do_one_initcall+0xa8/0x2f4)
(do_one_initcall) from [<80c0100c>] (kernel_init_freeable+0x178/0x1dc)
(kernel_init_freeable) from [<80807048>] (kernel_init+0x8/0x110)
(kernel_init) from [<80100114>] (ret_from_fork+0x14/0x20)
Additionally, the i2c_imx_isr() could wake up the wait queue
(imx_i2c_struct->queue) before its initialization happens.
The resource-managed framework should not be used for interrupt handling,
because the resource will be released too late - after disabling clocks.
The interrupt handler is not prepared for such case.
Fixes: 1c4b6c3bcf30 ("i2c: imx: implement bus recovery") Cc: <stable@vger.kernel.org> Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Acked-by: Oleksij Rempel <o.rempel@pengutronix.de> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The way the driver is implemented is buggy for the (admittedly unlikely)
use case where there are two RTCs with one having an interrupt configured
and the second not. This is caused by the fact that we use a global
rtc_class_ops struct which we modify depending on whether the irq number
is present or not.
Fix it by using two const ops structs with and without alarm operations.
While at it: not being able to request a configured interrupt is an error
so don't ignore it and bail out of probe().
Fix linkage error when CONFIG_BINFMT_ELF is selected but CONFIG_COREDUMP
is not:
ia64-linux-ld: arch/ia64/kernel/elfcore.o: in function `elf_core_write_extra_phdrs':
elfcore.c:(.text+0x172): undefined reference to `dump_emit'
ia64-linux-ld: arch/ia64/kernel/elfcore.o: in function `elf_core_write_extra_data':
elfcore.c:(.text+0x2b2): undefined reference to `dump_emit'
Fixes: 1fcccbac89f5 ("elf coredump: replace ELF_CORE_EXTRA_* macros by functions") Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/20200819064146.12529-1-krzk@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A detach hung is possible when a race occurs between the detach process
and the ubi background thread. The following sequences outline the race:
ubi thread: if (list_empty(&ubi->works)...
ubi detach: set_bit(KTHREAD_SHOULD_STOP, &kthread->flags)
=> by kthread_stop()
wake_up_process()
=> ubi thread is still running, so 0 is returned
ubi thread: set_current_state(TASK_INTERRUPTIBLE)
schedule()
=> ubi thread will never be scheduled again
ubi detach: wait_for_completion()
=> hung task!
To fix that, we need to check kthread_should_stop() after we set the
task state, so the ubi thread will either see the stop bit and exit or
the task state is reset to runnable such that it isn't scheduled out
indefinitely.
commit feb92d7d3813456c11dce21 "(ARC: perf: don't bail setup if pct irq
missing in device-tree)" introduced a silly brown-paper bag bug:
The assignment and comparison in an if statement were not bracketed
correctly leaving the order of evaluation undefined.
The problem is in the is_printable_array function that does not take the
zero byte into account and claim such string as not printable, so the
code will create byte array instead of string.
There is no need to dump authentication options while remounting,
because authentication initialization can only be doing once in
the first mount process. Dumping authentication mount options in
remount process may cause memory leak if UBIFS has already been
mounted with old authentication mount options.
When removing the last reference of an inode the size of an auth node
is already part of write_len. So we must not call ubifs_add_auth_dirt().
Call it only when needed.
Fix some potential memory leaks in error handling branches while
iterating xattr entries. For example, function ubifs_tnc_remove_ino()
forgets to free pxent if it exists. Similar problems also exist in
ubifs_purge_xattrs(), ubifs_add_orphan() and ubifs_jnl_write_inode().
Fix some potential memory leaks in error handling branches while
iterating dent entries. For example, function dbg_check_dir()
forgets to free pdent if it exists.
There's no protection in nfsd_dispatch() against a NULL .pc_func
helpers. A malicious NFS client can trigger a crash by invoking the
unused/unsupported NFSv2 ROOT or WRITECACHE procedures.
The current NFSD dispatcher does not support returning a void reply
to a non-NULL procedure, so the reply to both of these is wrong, for
the moment.
Cc: <stable@vger.kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since commit 0e0cb35b417f ("NFSv4: Handle NFS4ERR_OLD_STATEID in
CLOSE/OPEN_DOWNGRADE") the following livelock may occur if a CLOSE races
with the update of the nfs_state:
Process 1 Process 2 Server
========= ========= ========
OPEN file
OPEN file
Reply OPEN (1)
Reply OPEN (2)
Update state (1)
CLOSE file (1)
Reply OLD_STATEID (1)
CLOSE file (2)
Reply CLOSE (-1)
Update state (2)
wait for state change
OPEN file
wake
CLOSE file
OPEN file
wake
CLOSE file
...
...
We can avoid this situation by not issuing an immediate retry with a bumped
seqid when CLOSE/OPEN_DOWNGRADE receives NFS4ERR_OLD_STATEID. Instead,
take the same approach used by OPEN and wait at least 5 seconds for
outstanding stateid updates to complete if we can detect that we're out of
sequence.
Note that after this change it is still possible (though unlikely) that
CLOSE waits a full 5 seconds, bumps the seqid, and retries -- and that
attempt races with another OPEN at the same time. In order to avoid this
race (which would result in the livelock), update
nfs_need_update_open_stateid() to handle the case where:
- the state is NFS_OPEN_STATE, and
- the stateid doesn't match the current open stateid
Finally, nfs_need_update_open_stateid() is modified to be idempotent and
renamed to better suit the purpose of signaling that the stateid passed
is the next stateid in sequence.
Fixes: 0e0cb35b417f ("NFSv4: Handle NFS4ERR_OLD_STATEID in CLOSE/OPEN_DOWNGRADE") Cc: stable@vger.kernel.org # v5.4+ Signed-off-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
__get_user_atomic_128_aligned() stores to kaddr using stvx which is a
VMX store instruction, hence kaddr must be 16 byte aligned otherwise
the store won't occur as expected.
Unfortunately when we call __get_user_atomic_128_aligned() in
p9_hmi_special_emu(), the buffer we pass as kaddr (ie. vbuf) isn't
guaranteed to be 16B aligned. This means that the write to vbuf in
__get_user_atomic_128_aligned() has the bottom bits of the address
truncated. This results in other local variables being
overwritten. Also vbuf will not contain the correct data which results
in the userspace emulation being wrong and hence undetected user data
corruption.
In the past we've been mostly lucky as vbuf has ended up aligned but
this is fragile and isn't always true. CONFIG_STACKPROTECTOR in
particular can change the stack arrangement enough that our luck runs
out.
This issue only occurs on POWER9 Nimbus <= DD2.1 bare metal.
The fix is to align vbuf to a 16 byte boundary.
Fixes: 5080332c2c89 ("powerpc/64s: Add workaround for P9 vector CI load issue") Cc: stable@vger.kernel.org # v4.15+ Signed-off-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20201013043741.743413-1-mikey@neuling.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Every error log reported by OPAL is exported to userspace through a
sysfs interface and notified using kobject_uevent(). The userspace
daemon (opal_errd) then reads the error log and acknowledges the error
log is saved safely to disk. Once acknowledged the kernel removes the
respective sysfs file entry causing respective resources to be
released including kobject.
However it's possible the userspace daemon may already be scanning
elog entries when a new sysfs elog entry is created by the kernel.
User daemon may read this new entry and ack it even before kernel can
notify userspace about it through kobject_uevent() call. If that
happens then we have a potential race between
elog_ack_store->kobject_put() and kobject_uevent which can lead to
use-after-free of a kernfs object resulting in a kernel crash. eg:
BUG: Unable to handle kernel data access on read at 0x6b6b6b6b6b6b6bfb
Faulting instruction address: 0xc0000000008ff2a0
Oops: Kernel access of bad area, sig: 11 [#1]
LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA PowerNV
CPU: 27 PID: 805 Comm: irq/29-opal-elo Not tainted 5.9.0-rc2-gcc-8.2.0-00214-g6f56a67bcbb5-dirty #363
...
NIP kobject_uevent_env+0xa0/0x910
LR elog_event+0x1f4/0x2d0
Call Trace:
0x5deadbeef0000122 (unreliable)
elog_event+0x1f4/0x2d0
irq_thread_fn+0x4c/0xc0
irq_thread+0x1c0/0x2b0
kthread+0x1c4/0x1d0
ret_from_kernel_thread+0x5c/0x6c
This patch fixes this race by protecting the sysfs file
creation/notification by holding a reference count on kobject until we
safely send kobject_uevent().
The function create_elog_obj() returns the elog object which if used
by caller function will end up in use-after-free problem again.
However, the return value of create_elog_obj() function isn't being
used today and there is no need as well. Hence change it to return
void to make this fix complete.
Fixes: 774fea1a38c6 ("powerpc/powernv: Read OPAL error log and export it through sysfs") Cc: stable@vger.kernel.org # v3.15+ Reported-by: Oliver O'Halloran <oohall@gmail.com> Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Oliver O'Halloran <oohall@gmail.com> Reviewed-by: Vasant Hegde <hegdevasant@linux.vnet.ibm.com>
[mpe: Rework the logic to use a single return, reword comments, add oops] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20201006122051.190176-1-mpe@ellerman.id.au Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Similar to commit 89c140bbaeee ("pseries: Fix 64 bit logical memory block panic")
make sure different variables tracking lmb_size are updated to be 64 bit.
It's not done anything for a long time. Save the percpu variable, and
emit a warning to remind users to not expect it to do anything.
This uses pr_warn_once instead of pr_warn_ratelimit as testing
'ppc64_cpu --smt=off' on a 24 core / 4 SMT system showed the warning
to be noisy, as the online/offline loop is slow.
Fixes: 3fa8cad82b94 ("powerpc/pseries/cpuidle: smt-snooze-delay cleanup.") Cc: stable@vger.kernel.org # v3.14 Signed-off-by: Joel Stanley <joel@jms.id.au> Acked-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20200902000012.3440389-1-joel@jms.id.au Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A number of userspace utilities depend on making calls to RTAS to retrieve
information and update various things.
The existing API through which we expose RTAS to userspace exposes more
RTAS functionality than we actually need, through the sys_rtas syscall,
which allows root (or anyone with CAP_SYS_ADMIN) to make any RTAS call they
want with arbitrary arguments.
Many RTAS calls take the address of a buffer as an argument, and it's up to
the caller to specify the physical address of the buffer as an argument. We
allocate a buffer (the "RMO buffer") in the Real Memory Area that RTAS can
access, and then expose the physical address and size of this buffer in
/proc/powerpc/rtas/rmo_buffer. Userspace is expected to read this address,
poke at the buffer using /dev/mem, and pass an address in the RMO buffer to
the RTAS call.
However, there's nothing stopping the caller from specifying whatever
address they want in the RTAS call, and it's easy to construct a series of
RTAS calls that can overwrite arbitrary bytes (even without /dev/mem
access).
Additionally, there are some RTAS calls that do potentially dangerous
things and for which there are no legitimate userspace use cases.
In the past, this would not have been a particularly big deal as it was
assumed that root could modify all system state freely, but with Secure
Boot and lockdown we need to care about this.
We can't fundamentally change the ABI at this point, however we can address
this by implementing a filter that checks RTAS calls against a list
of permitted calls and forces the caller to use addresses within the RMO
buffer.
The list is based off the list of calls that are used by the librtas
userspace library, and has been tested with a number of existing userspace
RTAS utilities. For compatibility with any applications we are not aware of
that require other calls, the filter can be turned off at build time.
Cc: stable@vger.kernel.org Reported-by: Daniel Axtens <dja@axtens.net> Signed-off-by: Andrew Donnellan <ajd@linux.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20200820044512.7543-1-ajd@linux.ibm.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The sysfs function might race with stp_work_fn. To prevent that,
add the required locking. Another issue is that the sysfs functions
are checking the stp_online flag, but this flag just holds the user
setting whether STP is enabled. Add a flag to clock_sync_flag whether
stp_info holds valid data and use that instead.
Cc: stable@vger.kernel.org Signed-off-by: Sven Schnelle <svens@linux.ibm.com> Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com> Signed-off-by: Vasily Gorbik <gor@linux.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
caused by an attempt to free bootmem space that as from
commit b93ddc4f9156 ("mips: Reserve memory for the kernel image resources")
has not been anymore reserved due to the removal of generic MIPS arch code
that used to reserve all the memory from the beginning of RAM up to the
kernel load address.
This memory does need to be reserved on DEC platforms however as it is
used by REX firmware as working area, as per the TURBOchannel firmware
specification[1]:
Table 2-2 REX Memory Regions
-------------------------------------------------------------------------
Starting Ending
Region Address Address Use
-------------------------------------------------------------------------
0 0xa0000000 0xa000ffff Restart block, exception vectors,
REX stack and bss
1 0xa0010000 0xa0017fff Keyboard or tty drivers
2 0xa0018000 0xa001f3ff 1) CRT driver
3 0xa0020000 0xa002ffff boot, cnfg, init and t objects
4 0xa0020000 0xa002ffff 64KB scratch space
-------------------------------------------------------------------------
1) Note that the last 3 Kbytes of region 2 are reserved for backward
compatibility with previous system software.
-------------------------------------------------------------------------
(this table uses KSEG2 unmapped virtual addresses, which in the MIPS
architecture are offset from physical addresses by a fixed value of
0xa0000000 and therefore the regions referred do correspond to the
beginning of the physical address space) and we call into the firmware
on several occasions throughout the bootstrap process. It is believed
that pre-REX firmware used with non-TURBOchannel DEC platforms has the
same requirements, as hinted by note #1 cited.
Recreate the discarded reservation then, in DEC platform code, removing
the crash.
References:
[1] "TURBOchannel Firmware Specification", On-line version,
EK-TCAAD-FS-004, Digital Equipment Corporation, January 1993,
Chapter 2 "System Module Firmware", p. 2-5
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org> Fixes: b93ddc4f9156 ("mips: Reserve memory for the kernel image resources") Cc: stable@vger.kernel.org # v5.2+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Similar to commit 89c140bbaeee ("pseries: Fix 64 bit logical memory block panic")
make sure different variables tracking lmb_size are updated to be 64 bit.
One of a class of bugs pointed out by Lars in a recent review.
iio_push_to_buffers_with_timestamp assumes the buffer used is aligned
to the size of the timestamp (8 bytes). This is not guaranteed in
this driver which uses a 16 byte array of smaller elements on the stack.
This is fixed by using an explicit c structure. As there are no
holes in the structure, there is no possiblity of data leakage
in this case.
The explicit alignment of ts is not strictly necessary but potentially
makes the code slightly less fragile. It also removes the possibility
of this being cut and paste into another driver where the alignment
isn't already true.
Fixes: 36e0371e7764 ("iio:itg3200: Use iio_push_to_buffers_with_timestamp()") Reported-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: <Stable@vger.kernel.org> Link: https://lore.kernel.org/r/20200722155103.979802-6-jic23@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
One of a class of bugs pointed out by Lars in a recent review.
iio_push_to_buffers_with_timestamp assumes the buffer used is aligned
to the size of the timestamp (8 bytes). This is not guaranteed in
this driver which uses an array of smaller elements on the stack.
We move to a suitable structure in the iio_priv() data with alignment
explicitly requested. This data is allocated with kzalloc so no
data can leak apart from previous readings. Note that previously
no leak at all could occur, but previous readings should never
be a problem.
In this case the timestamp location depends on what other channels
are enabled. As such we can't use a structure without misleading
by suggesting only one possible timestamp location.
One of a class of bugs pointed out by Lars in a recent review.
iio_push_to_buffers_with_timestamp assumes the buffer used is aligned
to the size of the timestamp (8 bytes). This is not guaranteed in
this driver which uses an array of smaller elements on the stack.
We fix this issues by moving to a suitable structure in the iio_priv()
data with alignment explicitly requested. This data is allocated
with kzalloc so no data can leak apart from previous readings.
Note that previously no data could leak 'including' previous readings
but I don't think it is an issue to potentially leak them like
this now does.
In this case the postioning of the timestamp is depends on what
other channels are enabled. As such we cannot use a structure to
make the alignment explicit as it would be missleading by suggesting
only one possible location for the timestamp.
One of a class of bugs pointed out by Lars in a recent review.
iio_push_to_buffers_with_timestamp assumes the buffer used is aligned
to the size of the timestamp (8 bytes). This is not guaranteed in
this driver which uses a 24 byte array of smaller elements on the stack.
As Lars also noted this anti pattern can involve a leak of data to
userspace and that indeed can happen here. We close both issues by
moving to a suitable array in the iio_priv() data with alignment
explicitly requested. This data is allocated with kzalloc so no
data can leak appart from previous readings.
Depending on the enabled channels, the location of the timestamp
can be at various aligned offsets through the buffer. As such we
any use of a structure to enforce this alignment would incorrectly
suggest a single location for the timestamp. Comments adjusted to
express this clearly in the code.
Fixes: ac45e57f1590 ("iio: light: Add driver for Silabs si1132, si1141/2/3 and si1145/6/7 ambient light, uv index and proximity sensors") Reported-by: Lars-Peter Clausen <lars@metafoo.de> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Cc: Peter Meerwald-Stadler <pmeerw@pmeerw.net> Cc: <Stable@vger.kernel.org> Link: https://lore.kernel.org/r/20200722155103.979802-9-jic23@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The jz4780_dma_tx_status() function would check if a channel's cookie
state was set to 'completed', and if not, it would enter the critical
section. However, in that time frame, the jz4780_dma_chan_irq() function
was able to set the cookie to 'completed', and clear the jzchan->vchan
pointer, which was deferenced in the critical section of the first
function.
Fix this race by checking the channel's cookie state after entering the
critical function and not before.
Fixes: d894fc6046fe ("dmaengine: jz4780: add driver for the Ingenic JZ4780 DMA controller") Cc: stable@vger.kernel.org # v4.0 Signed-off-by: Paul Cercueil <paul@crapouillou.net> Reported-by: Artur Rojek <contact@artur-rojek.eu> Tested-by: Artur Rojek <contact@artur-rojek.eu> Link: https://lore.kernel.org/r/20201004140307.885556-1-paul@crapouillou.net Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
udf_process_sequence() allocates temporary array for processing
partition descriptors on volume which it fails to free. Free the array
when it is not needed anymore.
Fixes: 7b78fd02fb19 ("udf: Fix handling of Partition Descriptors") CC: stable@vger.kernel.org Reported-by: syzbot+128f4dd6e796c98b3760@syzkaller.appspotmail.com Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It has recently been reported that the "heartbeat" report from devices
like the 2nd-gen Intuos Pro (PTH-460, PTH-660, PTH-860) or the 2nd-gen
Bluetooth-enabled Intuos tablets (CTL-4100WL, CTL-6100WL) can cause the
driver to send a spurious BTN_TOUCH=0 once per second in the middle of
drawing. This can result in broken lines while drawing on Chrome OS.
The source of the issue has been traced back to a change which modified
the driver to only call `wacom_wac_pad_report()` once per report instead
of once per collection. As part of this change, pad-handling code was
removed from `wacom_wac_collection()` under the assumption that the
`WACOM_PEN_FIELD` and `WACOM_TOUCH_FIELD` checks would not be satisfied
when a pad or battery collection was being processed.
To be clear, the macros `WACOM_PAD_FIELD` and `WACOM_PEN_FIELD` do not
currently check exclusive conditions. In fact, most "pad" fields will
also appear to be "pen" fields simply due to their presence inside of
a Digitizer application collection. Because of this, the removal of
the check from `wacom_wac_collection()` just causes pad / battery
collections to instead trigger a call to `wacom_wac_pen_report()`
instead. The pen report function in turn resets the tip switch state
just prior to exiting, resulting in the observed BTN_TOUCH=0 symptom.
To correct this, we restore a version of the `WACOM_PAD_FIELD` check
in `wacom_wac_collection()` and return early. This effectively prevents
pad / battery collections from being reported until the very end of the
report as originally intended.
Both read-side users of func_table/func_buf need locking. Without that,
one can easily confuse the code by repeatedly setting altering strings
like:
while (1)
for (a = 0; a < 2; a++) {
struct kbsentry kbs = {};
strcpy((char *)kbs.kb_string, a ? ".\n" : "88888\n");
ioctl(fd, KDSKBSENT, &kbs);
}
When that program runs, one can get unexpected output by holding F1
(note the unxpected period on the last line):
.
88888
.8888
So protect all accesses to 'func_table' (and func_buf) by preexisting
'func_buf_lock'.
It is easy in 'k_fn' handler as 'puts_queue' is expected not to sleep.
On the other hand, KDGKBSENT needs a local (atomic) copy of the string
because copy_to_user can sleep. Use already allocated, but unused
'kbs->kb_string' for that purpose.
Note that the program above needs at least CAP_SYS_TTY_CONFIG.
This depends on the previous patch and on the func_buf_lock lock added
in commit 46ca3f735f34 (tty/vt: fix write/write race in ioctl(KDSKBSENT)
handler) in 5.2.
Use 'strlen' of the string, add one for NUL terminator and simply do
'copy_to_user' instead of the explicit 'for' loop. This makes the
KDGKBSENT case more compact.
The only thing we need to take care about is NULL 'func_table[i]'. Use
an empty string in that case.
The original check for overflow could never trigger as the func_buf
strings are always shorter or equal to 'struct kbsentry's.
If i915.ko is being used as a passthrough device, it does not know if
the host is using intel_iommu. Mixing the iommu and gfx causes a few
issues (such as scanout overfetch) which we need to workaround inside
the driver, so if we detect we are running under a hypervisor, also
assume the device access is being virtualised.
Reported-by: Stefan Fritsch <sf@sfritsch.de> Suggested-by: Stefan Fritsch <sf@sfritsch.de> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Stefan Fritsch <sf@sfritsch.de> Cc: stable@vger.kernel.org Tested-by: Stefan Fritsch <sf@sfritsch.de> Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20201019101523.4145-1-chris@chris-wilson.co.uk
(cherry picked from commit f566fdcd6cc49a9d5b5d782f56e3e7cb243f01b8) Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fsl_usb2_device_register() should stop init if dma_set_mask() return
error.
Fixes: cae058610465 ("drivers/usb/host: fsl: Set DMA_MASK of usb platform device") Reviewed-by: Peter Chen <peter.chen@nxp.com> Signed-off-by: Ran Wang <ran.wang_1@nxp.com> Link: https://lore.kernel.org/r/20201010060308.33693-1-ran.wang_1@nxp.com Cc: stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Current tcpm_detach() only reset hard_reset_count if port->attached
is true, this may cause this counter clear is missed if the CC
disconnect event is generated after tcpm_port_reset() is done
by other events, e.g. VBUS off comes first before CC disconect for
a power sink, in that case the first tcpm_detach() will only clear
port->attached flag but leave hard_reset_count there because
tcpm_port_is_disconnected() is still false, then later tcpm_detach()
by CC disconnect will directly return due to port->attached is cleared,
finally this will result tcpm will not try hard reset or error recovery
for later attach.
ChiYuan reported this issue on his platform with below tcpm trace:
After power sink session setup after hard reset 2 times, detach
from the power source and then attach:
[ 4848.046358] VBUS off
[ 4848.046384] state change SNK_READY -> SNK_UNATTACHED
[ 4848.050908] Setting voltage/current limit 0 mV 0 mA
[ 4848.050936] polarity 0
[ 4848.052593] Requesting mux state 0, usb-role 0, orientation 0
[ 4848.053222] Start toggling
[ 4848.086500] state change SNK_UNATTACHED -> TOGGLING
[ 4848.089983] CC1: 0 -> 0, CC2: 3 -> 3 [state TOGGLING, polarity 0, connected]
[ 4848.089993] state change TOGGLING -> SNK_ATTACH_WAIT
[ 4848.090031] pending state change SNK_ATTACH_WAIT -> SNK_DEBOUNCED @200 ms
[ 4848.141162] CC1: 0 -> 0, CC2: 3 -> 0 [state SNK_ATTACH_WAIT, polarity 0, disconnected]
[ 4848.141170] state change SNK_ATTACH_WAIT -> SNK_ATTACH_WAIT
[ 4848.141184] pending state change SNK_ATTACH_WAIT -> SNK_UNATTACHED @20 ms
[ 4848.163156] state change SNK_ATTACH_WAIT -> SNK_UNATTACHED [delayed 20 ms]
[ 4848.163162] Start toggling
[ 4848.216918] CC1: 0 -> 0, CC2: 0 -> 3 [state TOGGLING, polarity 0, connected]
[ 4848.216954] state change TOGGLING -> SNK_ATTACH_WAIT
[ 4848.217080] pending state change SNK_ATTACH_WAIT -> SNK_DEBOUNCED @200 ms
[ 4848.231771] CC1: 0 -> 0, CC2: 3 -> 0 [state SNK_ATTACH_WAIT, polarity 0, disconnected]
[ 4848.231800] state change SNK_ATTACH_WAIT -> SNK_ATTACH_WAIT
[ 4848.231857] pending state change SNK_ATTACH_WAIT -> SNK_UNATTACHED @20 ms
[ 4848.256022] state change SNK_ATTACH_WAIT -> SNK_UNATTACHED [delayed20 ms]
[ 4848.256049] Start toggling
[ 4848.871148] VBUS on
[ 4848.885324] CC1: 0 -> 0, CC2: 0 -> 3 [state TOGGLING, polarity 0, connected]
[ 4848.885372] state change TOGGLING -> SNK_ATTACH_WAIT
[ 4848.885548] pending state change SNK_ATTACH_WAIT -> SNK_DEBOUNCED @200 ms
[ 4849.088240] state change SNK_ATTACH_WAIT -> SNK_DEBOUNCED [delayed200 ms]
[ 4849.088284] state change SNK_DEBOUNCED -> SNK_ATTACHED
[ 4849.088291] polarity 1
[ 4849.088769] Requesting mux state 1, usb-role 2, orientation 2
[ 4849.088895] state change SNK_ATTACHED -> SNK_STARTUP
[ 4849.088907] state change SNK_STARTUP -> SNK_DISCOVERY
[ 4849.088915] Setting voltage/current limit 5000 mV 0 mA
[ 4849.088927] vbus=0 charge:=1
[ 4849.090505] state change SNK_DISCOVERY -> SNK_WAIT_CAPABILITIES
[ 4849.090828] pending state change SNK_WAIT_CAPABILITIES -> SNK_READY @240 ms
[ 4849.335878] state change SNK_WAIT_CAPABILITIES -> SNK_READY [delayed240 ms]
this patch fix this issue by clear hard_reset_count at any cases
of cc disconnect, í.e. don't check port->attached flag.
Fixes: 4b4e02c83167 ("typec: tcpm: Move out of staging") Cc: stable@vger.kernel.org Reported-and-tested-by: ChiYuan Huang <cy_huang@richtek.com> Reviewed-by: Guenter Roeck <linux@roeck-us.net> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Signed-off-by: Li Jun <jun.li@nxp.com> Link: https://lore.kernel.org/r/1602500592-3817-1-git-send-email-jun.li@nxp.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
is incorrect in case of a delayed work and causes warnings, usually from
the workqueue:
> WARNING: CPU: 0 PID: 0 at kernel/workqueue.c:1474 __queue_work+0x480/0x528.
When this happens, USB eventually stops working completely after a while.
Also the ACM_ERROR_DELAY bit is never set, so the cooldown mechanism
previously introduced cannot be triggered and acm_submit_read_urb() is
never called.
This changes makes the cdc-acm driver use a single delayed work, fixing the
pointer arithmetic in acm_softint() and set the ACM_ERROR_DELAY when the
cooldown mechanism appear to be needed.
According the programming guide (for all DWC3 IPs), when the driver
handles ClearFeature(halt) request, it should issue CLEAR_STALL command
_after_ the END_TRANSFER command completes. The END_TRANSFER command may
take some time to complete. So, delay the ClearFeature(halt) request
control status stage and wait for END_TRANSFER command completion
interrupt. Only after END_TRANSFER command completes that the driver
may issue CLEAR_STALL command.
The function driver may queue new requests right after halting the
endpoint (i.e. queue new requests while the endpoint is stalled).
There's no restriction preventing it from doing so. However, dwc3
currently drops those requests after CLEAR_STALL. The driver should only
drop started requests. Keep the pending requests in the pending list to
resume and process them after the host issues ClearFeature(Halt) to the
endpoint.
No need to trigger runtime pm in driver removal, otherwise if user
disable auto suspend via sys file, runtime suspend may be entered,
which will call dwc3_core_exit() again and there will be clock disable
not balance warning:
When preparing for SG, not all the entries are prepared at once. When
resume, don't use the remaining request length to calculate for MPS
alignment. Use the entire request->length to do that.
Cc: stable@vger.kernel.org Fixes: 5d187c0454ef ("usb: dwc3: gadget: Don't setup more than requested") Signed-off-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com> Signed-off-by: Felipe Balbi <balbi@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current ZLP handling for ep0 requests is only for control IN
requests. For OUT direction, DWC3 needs to check and setup for MPS
alignment.
Usually, control OUT requests can indicate its transfer size via the
wLength field of the control message. So usb_request->zero is usually
not needed for OUT direction. To handle ZLP OUT for control endpoint,
make sure the TRB is MPS size.
Similar to some other IA platforms, Elkhart Lake too depends on the
PMU register write to request transition of Dx power state.
Thus, we add the PCI_DEVICE_ID_INTEL_EHLLP to the list of devices that
shall execute the ACPI _DSM method during D0/D3 sequence.
[heikki.krogerus@linux.intel.com: included Fixes tag]
Fixes: dbb0569de852 ("usb: dwc3: pci: Add Support for Intel Elkhart Lake Devices") Cc: stable@vger.kernel.org Signed-off-by: Raymond Tan <raymond.tan@intel.com> Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Signed-off-by: Felipe Balbi <balbi@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Very sporadically I had test case btrfs/069 from fstests hanging (for
years, it is not a recent regression), with the following traces in
dmesg/syslog:
This was weird because the stack traces show that a transaction commit,
triggered by a device replace operation, is blocking trying to pause any
running scrubs but there are no stack traces of blocked tasks doing a
scrub.
After poking around with drgn, I noticed there was a scrub task that was
constantly running and blocking for shorts periods of time:
while (atomic_read(&rc->elems)) {
if (!atomic_read(&fs_info->reada_works_cnt))
reada_start_machine(fs_info);
wait_event_timeout(rc->wait, atomic_read(&rc->elems) == 0,
(HZ + 9) / 10);
}
(...)
So the counter "rc->elems" was set to 1 and never decreased to 0, causing
the scrub task to loop forever in that function. Then I used the following
script for drgn to check the readahead requests:
def dump_re(re):
nzones = re.nzones.value_()
print(f're at {hex(re.value_())}')
print(f'\t logical {re.logical.value_()}')
print(f'\t refcnt {re.refcnt.value_()}')
print(f'\t nzones {nzones}')
for i in range(nzones):
dev = re.zones[i].device
name = dev.name.str.string_()
print(f'\t\t dev id {dev.devid.value_()} name {name}')
print()
for _, e in radix_tree_for_each(fs_info.reada_tree):
re = cast('struct reada_extent *', e)
dump_re(re)
$ drgn dump_reada.py
re at 0xffff8f3da9d25ad8
logical 38928384
refcnt 1
nzones 1
dev id 0 name b'/dev/sdd'
$
So there was one readahead extent with a single zone corresponding to the
source device of that last device replace operation logged in dmesg/syslog.
Also the ID of that zone's device was 0 which is a special value set in
the source device of a device replace operation when the operation finishes
(constant BTRFS_DEV_REPLACE_DEVID set at btrfs_dev_replace_finishing()),
confirming again that device /dev/sdd was the source of a device replace
operation.
Normally there should be as many zones in the readahead extent as there are
devices, and I wasn't expecting the extent to be in a block group with a
'single' profile, so I went and confirmed with the following drgn script
that there weren't any single profile block groups:
def bg_flags_string(bg):
flags = bg.flags.value_()
ret = ''
if flags & BTRFS_BLOCK_GROUP_DATA:
ret = 'data'
if flags & BTRFS_BLOCK_GROUP_METADATA:
if len(ret) > 0:
ret += '|'
ret += 'meta'
if flags & BTRFS_BLOCK_GROUP_SYSTEM:
if len(ret) > 0:
ret += '|'
ret += 'system'
if flags & BTRFS_BLOCK_GROUP_RAID0:
ret += ' raid0'
elif flags & BTRFS_BLOCK_GROUP_RAID1:
ret += ' raid1'
elif flags & BTRFS_BLOCK_GROUP_DUP:
ret += ' dup'
elif flags & BTRFS_BLOCK_GROUP_RAID10:
ret += ' raid10'
elif flags & BTRFS_BLOCK_GROUP_RAID5:
ret += ' raid5'
elif flags & BTRFS_BLOCK_GROUP_RAID6:
ret += ' raid6'
elif flags & BTRFS_BLOCK_GROUP_RAID1C3:
ret += ' raid1c3'
elif flags & BTRFS_BLOCK_GROUP_RAID1C4:
ret += ' raid1c4'
else:
ret += ' single'
return ret
def dump_bg(bg):
print()
print(f'block group at {hex(bg.value_())}')
print(f'\t start {bg.start.value_()} length {bg.length.value_()}')
print(f'\t flags {bg.flags.value_()} - {bg_flags_string(bg)}')
bg_root = fs_info.block_group_cache_tree.address_of_()
for bg in rbtree_inorder_for_each_entry('struct btrfs_block_group', bg_root, 'cache_node'):
dump_bg(bg)
$ drgn dump_block_groups.py
block group at 0xffff8f3d673b0400
start 22020096 length 16777216
flags 258 - system raid6
block group at 0xffff8f3d53ddb400
start 38797312 length 536870912
flags 260 - meta raid6
block group at 0xffff8f3d5f4d9c00
start 575668224 length 2147483648
flags 257 - data raid6
block group at 0xffff8f3d08189000
start 2723151872 length 67108864
flags 258 - system raid6
block group at 0xffff8f3db70ff000
start 2790260736 length 1073741824
flags 260 - meta raid6
block group at 0xffff8f3d5f4dd800
start 3864002560 length 67108864
flags 258 - system raid6
block group at 0xffff8f3d67037000
start 3931111424 length 2147483648
flags 257 - data raid6
$
So there were only 2 reasons left for having a readahead extent with a
single zone: reada_find_zone(), called when creating a readahead extent,
returned NULL either because we failed to find the corresponding block
group or because a memory allocation failed. With some additional and
custom tracing I figured out that on every further ocurrence of the
problem the block group had just been deleted when we were looping to
create the zones for the readahead extent (at reada_find_extent()), so we
ended up with only one zone in the readahead extent, corresponding to a
device that ends up getting replaced.
So after figuring that out it became obvious why the hang happens:
1) Task A starts a scrub on any device of the filesystem, except for
device /dev/sdd;
2) Task B starts a device replace with /dev/sdd as the source device;
3) Task A calls btrfs_reada_add() from scrub_stripe() and it is currently
starting to scrub a stripe from block group X. This call to
btrfs_reada_add() is the one for the extent tree. When btrfs_reada_add()
calls reada_add_block(), it passes the logical address of the extent
tree's root node as its 'logical' argument - a value of 38928384;
4) Task A then enters reada_find_extent(), called from reada_add_block().
It finds there isn't any existing readahead extent for the logical
address 38928384, so it proceeds to the path of creating a new one.
It calls btrfs_map_block() to find out which stripes exist for the block
group X. On the first iteration of the for loop that iterates over the
stripes, it finds the stripe for device /dev/sdd, so it creates one
zone for that device and adds it to the readahead extent. Before getting
into the second iteration of the loop, the cleanup kthread deletes block
group X because it was empty. So in the iterations for the remaining
stripes it does not add more zones to the readahead extent, because the
calls to reada_find_zone() returned NULL because they couldn't find
block group X anymore.
As a result the new readahead extent has a single zone, corresponding to
the device /dev/sdd;
4) Before task A returns to btrfs_reada_add() and queues the readahead job
for the readahead work queue, task B finishes the device replace and at
btrfs_dev_replace_finishing() swaps the device /dev/sdd with the new
device /dev/sdg;
5) Task A returns to reada_add_block(), which increments the counter
"->elems" of the reada_control structure allocated at btrfs_reada_add().
Then it returns back to btrfs_reada_add() and calls
reada_start_machine(). This queues a job in the readahead work queue to
run the function reada_start_machine_worker(), which calls
__reada_start_machine().
At __reada_start_machine() we take the device list mutex and for each
device found in the current device list, we call
reada_start_machine_dev() to start the readahead work. However at this
point the device /dev/sdd was already freed and is not in the device
list anymore.
This means the corresponding readahead for the extent at 38928384 is
never started, and therefore the "->elems" counter of the reada_control
structure allocated at btrfs_reada_add() never goes down to 0, causing
the call to btrfs_reada_wait(), done by the scrub task, to wait forever.
Note that the readahead request can be made either after the device replace
started or before it started, however in pratice it is very unlikely that a
device replace is able to start after a readahead request is made and is
able to complete before the readahead request completes - maybe only on a
very small and nearly empty filesystem.
This hang however is not the only problem we can have with readahead and
device removals. When the readahead extent has other zones other than the
one corresponding to the device that is being removed (either by a device
replace or a device remove operation), we risk having a use-after-free on
the device when dropping the last reference of the readahead extent.
For example if we create a readahead extent with two zones, one for the
device /dev/sdd and one for the device /dev/sde:
1) Before the readahead worker starts, the device /dev/sdd is removed,
and the corresponding btrfs_device structure is freed. However the
readahead extent still has the zone pointing to the device structure;
2) When the readahead worker starts, it only finds device /dev/sde in the
current device list of the filesystem;
3) It starts the readahead work, at reada_start_machine_dev(), using the
device /dev/sde;
4) Then when it finishes reading the extent from device /dev/sde, it calls
__readahead_hook() which ends up dropping the last reference on the
readahead extent through the last call to reada_extent_put();
5) At reada_extent_put() it iterates over each zone of the readahead extent
and attempts to delete an element from the device's 'reada_extents'
radix tree, resulting in a use-after-free, as the device pointer of the
zone for /dev/sdd is now stale. We can also access the device after
dropping the last reference of a zone, through reada_zone_release(),
also called by reada_extent_put().
And a device remove suffers the same problem, however since it shrinks the
device size down to zero before removing the device, it is very unlikely to
still have readahead requests not completed by the time we free the device,
the only possibility is if the device has a very little space allocated.
While the hang problem is exclusive to scrub, since it is currently the
only user of btrfs_reada_add() and btrfs_reada_wait(), the use-after-free
problem affects any path that triggers readhead, which includes
btree_readahead_hook() and __readahead_hook() (a readahead worker can
trigger readahed for the children of a node) for example - any path that
ends up calling reada_add_block() can trigger the use-after-free after a
device is removed.
So fix this by waiting for any readahead requests for a device to complete
before removing a device, ensuring that while waiting for existing ones no
new ones can be made.
This problem has been around for a very long time - the readahead code was
added in 2011, device remove exists since 2008 and device replace was
introduced in 2013, hard to pick a specific commit for a git Fixes tag.
CC: stable@vger.kernel.org # 4.4+ Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If we fail to find suitable zones for a new readahead extent, we end up
leaving a stale pointer in the global readahead extents radix tree
(fs_info->reada_tree), which can trigger the following trace later on:
1) At reada_find_extent() we don't find any existing readahead extent for
the metadata extent starting at logical address X;
2) So we proceed to create a new one. We then call btrfs_map_block() to get
information about which stripes contain extent X;
3) After that we iterate over the stripes and create only one zone for the
readahead extent - only one because reada_find_zone() returned NULL for
all iterations except for one, either because a memory allocation failed
or it couldn't find the block group of the extent (it may have just been
deleted);
4) We then add the new readahead extent to the readahead extents radix
tree at fs_info->reada_tree;
5) Then we iterate over each zone of the new readahead extent, and find
that the device used for that zone no longer exists, because it was
removed or it was the source device of a device replace operation.
Since this left 'have_zone' set to 0, after finishing the loop we jump
to the 'error' label, call kfree() on the new readahead extent and
return without removing it from the radix tree at fs_info->reada_tree;
6) Any future call to reada_find_extent() for the logical address X will
find the stale pointer in the readahead extents radix tree, increment
its reference counter, which can trigger the use-after-free right
away or return it to the caller reada_add_block() that results in the
use-after-free of the example trace above.
So fix this by making sure we delete the readahead extent from the radix
tree if we fail to setup zones for it (when 'have_zone = 0').
Fixes: 319450211842ba ("btrfs: reada: bypass adding extent when all zone failed") CC: stable@vger.kernel.org # 4.9+ Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / users / dwmw2 / linux.git / log