GCC assumes that stack is aligned to 16-byte on call sites [1].
Since GCC 8, GCC began using 16-byte aligned SSE instructions to
implement assignments to structs on stack. When
CC_OPTIMIZE_FOR_PERFORMANCE is enabled, this affects
os-Linux/sigio.c, write_sigio_thread:
struct pollfds *fds, tmp;
tmp = current_poll;
Note that struct pollfds is exactly 16 bytes in size.
GCC 8+ generates assembly similar to:
movdqa (%rdi),%xmm0
movaps %xmm0,-0x50(%rbp)
This is an issue, because movaps will #GP if -0x50(%rbp) is not
aligned to 16 bytes [2], and how rbp gets assigned to is via glibc
clone thread_start, then function prologue, going though execution
trace similar to (showing only relevant instructions):
sub $0x10,%rsi
mov %rcx,0x8(%rsi)
mov %rdi,(%rsi)
syscall
pop %rax
pop %rdi
callq *%rax
push %rbp
mov %rsp,%rbp
The stack pointer always points to the topmost element on stack,
rather then the space right above the topmost. On push, the
pointer decrements first before writing to the memory pointed to
by it. Therefore, there is no need to have the stack pointer
pointer always point to valid memory unless the stack is poped;
so the `- sizeof(void *)` in the code is unnecessary.
On the other hand, glibc reserves the 16 bytes it needs on stack
and pops itself, so by the call instruction the stack pointer
is exactly the caller-supplied sp. It then push the 16 bytes of
the return address and the saved stack pointer, so the base
pointer will be 16-byte aligned if and only if the caller
supplied sp is 16-byte aligned. Therefore, the caller must supply
a 16-byte aligned pointer, which `stack + UM_KERN_PAGE_SIZE`
already satisfies.
On a side note, musl is unaffected by this issue because it forces
16 byte alignment via `and $-16,%rsi` in its clone wrapper.
Similarly, glibc i386 is also unaffected because it has
`andl $0xfffffff0, %ecx`.
To reproduce this bug, enable CONFIG_UML_RTC and
CC_OPTIMIZE_FOR_PERFORMANCE. uml_rtc will call
add_sigio_fd which will then cause write_sigio_thread to either go
into segfault loop or panic with "Segfault with no mm".
Similarly, signal stacks will be aligned by the host kernel upon
signal delivery. `- sizeof(void *)` to sigaltstack is
unconventional and extraneous.
On a related note, initialization of longjmp buffers do require
`- sizeof(void *)`. This is to account for the return address
that would have been pushed to the stack at the call site.
The reason for uml to respect 16-byte alignment, rather than
telling GCC to assume 8-byte alignment like the host kernel since
commit d9b0cde91c ("x86-64, gcc: Use
-mpreferred-stack-boundary=3 if supported"), is because uml links
against libc. There is no reason to assume libc is also compiled
with that flag and assumes 8-byte alignment rather than 16-byte.
[1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=40838
[2] https://c9x.me/x86/html/file_module_x86_id_180.html
Signed-off-by: YiFei Zhu <zhuyifei1999@gmail.com>
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Richard Weinberger <richard@nod.at>
vmalloc() heavy workloads in UML are extremely slow, due to
flushing the entire kernel VM space (flush_tlb_kernel_vm())
on the first segfault.
Implement flush_cache_vmap() to avoid that, and while at it
also add flush_cache_vunmap() since it's trivial.
This speeds up my vmalloc() heavy test of copying files out
from /sys/kernel/debug/gcov/ by 30x (from 30s to 1s.)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-By: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
The generic asm futex implementation emulates atomic access to
memory by doing a get_user followed by put_user. These translate
to two mapping operations on UML with paging enabled in the
meantime. This, in turn may end up changing interrupts,
invoking the signal loop, etc.
This replaces the generic implementation by a mapping followed
by an operation on the mapped segment.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This patch enables the use of optimized xor routines from the x86
tree as well as the necessary fpu api shims so they can work on
UML.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
1. Reflect host cpu flags into the UML instance so they can
be used to select the correct implementations for xor, crypto, etc.
2. Reflect host cache alignment into UML instance. This is
important when running 32 bit on a 64 bit host as 32 bit by
default aligns to 32 while the actual alignment should be 64.
Ditto for some Xeons which align at 128.
Signed-off-by: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This tag contains a few small fixes, allows to build the Berlin reset
driver as a module, and adds stubs to the reset controller API to allow
compile-testing drivers outside of drivers/reset without enabling the
reset framework.
-----BEGIN PGP SIGNATURE-----
iI0EABYIADUWIQRRO6F6WdpH1R0vGibVhaclGDdiwAUCYMtFIxcccC56YWJlbEBw
ZW5ndXRyb25peC5kZQAKCRDVhaclGDdiwEIsAP4m0pQOBknwAy8IpNXhXzqcXd3d
fcFn36QGzMxA3CEI7QD+IQMxfKvCOZNUVKaDQa3azX6gvbooqsr8xl+5MYsHSQU=
=/q1G
-----END PGP SIGNATURE-----
gpgsig -----BEGIN PGP SIGNATURE-----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=AwL+
-----END PGP SIGNATURE-----
Merge tag 'reset-for-v5.14-2' of git://git.pengutronix.de/pza/linux into arm/drivers
Reset controller updates for v5.14, part2
This tag contains a few small fixes, allows to build the Berlin reset
driver as a module, and adds stubs to the reset controller API to allow
compile-testing drivers outside of drivers/reset without enabling the
reset framework.
* tag 'reset-for-v5.14-2' of git://git.pengutronix.de/pza/linux:
reset: Add compile-test stubs
reset: berlin: support module build
reset: bail if try_module_get() fails
reset: mchp: sparx5: fix return value check in mchp_sparx5_map_io()
reset: lantiq: use devm_reset_controller_register()
reset: hi6220: Use the correct HiSilicon copyright
Link: https://lore.kernel.org/r/14d33ac19b2a107e97ce1ab264987b707baa9ba7.camel@pengutronix.de
Signed-off-by: Olof Johansson <olof@lixom.net>
There doesn't seem to be any reason for the rpath being set in
the binaries, at on systems that I tested on. On the other hand,
setting rpath is actually harming binaries in some cases, e.g.
if using nix-based compilation environments where /lib & /lib64
are not part of the actual environment.
Add a new Kconfig option (under EXPERT, for less user confusion)
that allows disabling the rpath additions.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Sphinx 3.0 works at this point (albeit slowly) so stop scaring people
with a loud warning. We also don't need to babble about CJK support in the
LaTeX build.
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
The UM virtual PCI devices currently cannot be suspended properly
since the virtio driver already disables VQs well before the PCI
bus's suspend_noirq wants to complete the transition by writing to
PCI config space.
After trying around for a long time with moving the devices on the
DPM list, trying to create dependencies between them, etc. I gave
up and instead added UML specific cross-driver API that lets the
virt-pci code enable not suspending/resuming VQs for its devices.
This then allows the PCI bus suspend_noirq to still talk to the
device, and suspend/resume works properly.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
To support testing of PCI/PCIe drivers in UML, add a PCI bus
support driver. This driver uses virtio, which in UML is really
just vhost-user, to talk to devices, and adds the devices to
the virtual PCI bus in the system.
Since virtio already allows DMA/bus mastering this really isn't
all that hard, of course we need the logic_iomem infrastructure
that was added by a previous patch.
The protocol to talk to the device is has a few fairly simple
messages for reading to/writing from config and IO spaces, and
messages for the device to send the various interrupts (INT#,
MSI/MSI-X and while suspended PME#).
Note that currently no offical virtio device ID is assigned for
this protocol, as a consequence this patch requires defining it
in the Kconfig, with a default that makes the driver refuse to
work at all.
Finally, in order to add support for MSI/MSI-X interrupts, some
small changes are needed in the UML IRQ code, it needs to have
more interrupts, changing NR_IRQS from 64 to 128 if this driver
is enabled, but not actually use them for anything so that the
generic IRQ domain/MSI infrastructure can allocate IRQ numbers.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
If we happen to get multiple messages while IRQS are already
suspended, we still need to handle them, since otherwise the
simulation blocks.
Remove the "prevent nesting" part, time_travel_add_irq_event()
will deal with being called multiple times just fine.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
We should be able to ndelay() from any context, even from an
interrupt context! However, this is broken (not functionally,
but locking-wise) in time-travel because we'll get into the
time-travel code and enable interrupts to handle messages on
other time-travel aware subsystems (only virtio for now).
Luckily, I've already reworked the time-travel aware signal
(interrupt) delivery for suspend/resume to have a time travel
handler, which runs directly in the context of the signal and
not from the Linux interrupt.
In order to fix this time-travel issue then, we need to do a
few things:
1) rework the signal handling code to call time-travel handlers
(only) if interrupts are disabled but signals aren't blocked,
instead of marking it only pending there. This is needed to
not deadlock other communication.
2) rework time-travel to not enable interrupts while it's
waiting for a message;
3) rework time-travel to not (just) disable interrupts but
rather block signals at a lower level while it needs them
disabled for communicating with the controller.
Finally, since now we can actually spend even virtual time
in interrupts-disabled sections, the delay warning when we
deliver a time-travel delayed interrupt is no longer valid,
things can (and should) now get delayed.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This will be necessary in the userspace side to fix the
signal/interrupt handling in time-travel=ext mode, which
is the next patch.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Use signals_enabled instead of always jumping through
a function call to read it, there's not much point in
that.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This function doesn't exist, remove its declaration.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-By: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Add IO memory emulation that uses callbacks for read/write to
the allocated regions. The callbacks can be registered by the
users using logic_iomem_alloc().
To use, an architecture must 'select LOGIC_IOMEM' in Kconfig
and then include <asm-generic/logic_io.h> into asm/io.h to get
the __raw_read*/__raw_write* functions.
Optionally, an architecture may 'select LOGIC_IOMEM_FALLBACK'
in which case non-emulated regions will 'fall back' to the
various real_* functions that must then be provided.
Cc: Arnd Bergmann <arnd@kernel.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Adjust the kconfig a little to allow disabling NO_IOMEM in UML. To
make an "allyesconfig" with CONFIG_NO_IOMEM=n build, adjust a few
Kconfig things elsewhere and add dummy asm/fb.h and asm/vga.h files.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Mauro says:
> As discussed at:
> https://lore.kernel.org/linux-doc/871r9k6rmy.fsf@meer.lwn.net/
>
> It is better to avoid using :doc:`foo` to refer to Documentation/foo.rst, as the
> automarkup.py extension should handle it automatically, on most cases.
>
> There are a couple of exceptions to this rule:
>
> 1. when :doc: tag is used to point to a kernel-doc DOC: markup;
> 2. when it is used with a named tag, e. g. :doc:`some name <foo>`;
>
> On this series:
>
> Patch 1 manually adjust the references inside driver-api/pm/devices.rst,
> as there it uses :file:`foo` to refer to some Documentation/ files;
>
> Patch 2 converts a table at Documentation/dev-tools/kunit/api/index.rst
> into a list, carefully avoiding the
>
> The remaining patches convert the other occurrences via a replace script.
> They were manually edited, in order to honour 80-columns where possible.
>
> This series based on docs-next branch. In order to avoid merge conflicts,
> I rebased it internally against yesterday's linux-next, dropping a patch
> and a hunk that would have caused conflicts there.
>
The :doc:`foo` tag is auto-generated via automarkup.py.
So, use the filename at the sources, instead of :doc:`foo`.
Reviewed-by: David Gow <davidgow@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/6fde409079959a95b62b9b2692503608d7ff0dbd.1623824363.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
We'll be replacing :doc:`foo` references to
Documentation/foo.rst. Yet, here it happens inside a table.
Doing a search-and-replace would break it.
Yet, as there's no good reason to use a table there,
let's just convert it into a list.
Reviewed-by: David Gow <davidgow@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/07d3a8ccafbb6345d6e78fb090290859e84361a1.1623824363.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
The conversion tools used during DocBook/LaTeX/html/Markdown->ReST
conversion and some cut-and-pasted text contain some characters that
aren't easily reachable on standard keyboards and/or could cause
troubles when parsed by the documentation build system.
Replace the occurences of the following characters:
- U+00a0 (' '): NO-BREAK SPACE
as it can cause lines being truncated on PDF output
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/8036126a59adb720dbc9233341ad5a08531cf73f.1623826294.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
The conversion tools used during DocBook/LaTeX/html/Markdown->ReST
conversion and some cut-and-pasted text contain some characters that
aren't easily reachable on standard keyboards and/or could cause
troubles when parsed by the documentation build system.
Replace the occurences of the following characters:
- U+00a0 (' '): NO-BREAK SPACE
as it can cause lines being truncated on PDF output
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/9bd9f5c067c4b068a974730a14fe8d68e1be0c9a.1623826294.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
The conversion tools used during DocBook/LaTeX/html/Markdown->ReST
conversion and some cut-and-pasted text contain some characters that
aren't easily reachable on standard keyboards and/or could cause
troubles when parsed by the documentation build system.
Replace the occurences of the following characters:
- U+2217 ('∗'): ASTERISK OPERATOR
use ASCII asterisk instead of the ASTERISK OPERATOR
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Acked-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/r/c5c3c384c48779ca7c9dcd90183cefe20ac82928.1623826294.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>