The NVlink IOMMU group setup is only relevant to NVLink devices so move
it into the NPU containment zone. This let us remove some prototypes in
pci.h and staticfy some function definitions.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-8-oohall@gmail.com
Move it in with the rest of the TCE wrangling rather than carting around
a static prototype in pci-ioda.c
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-7-oohall@gmail.com
Historically adding devices to their respective iommu group has been
handled by the post-init phb fixup for most devices. This was done
because:
1) The IOMMU group is tied to the PE (usually) so we can only setup the
iommu groups after we've done resource allocation since BAR location
determines the device's PE, and:
2) The sysfs directory for the pci_dev needs to be available since
iommu_add_device() wants to add an attribute for the iommu group.
However, since commit 30d87ef8b3 ("powerpc/pci: Fix
pcibios_setup_device() ordering") both conditions are met when
hose->ops->dma_dev_setup() is called so there's no real need to do
this in the fixup.
Moving the call to iommu_add_device() into pnv_pci_ioda_dma_setup_dev()
is a nice cleanup since it puts all the per-device IOMMU setup into one
place. It also results in all (non-nvlink) devices getting their iommu
group via a common path rather than relying on the bus notifier hack
in pnv_tce_iommu_bus_notifier() to handle the adding VFs and
hotplugged devices to their group.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-5-oohall@gmail.com
Move the registration of IOMMU groups out of the post-phb init fixup and
into when we configure DMA for a PE. For most devices this doesn't
result in any functional changes, but for NVLink attached GPUs it
requires a bit of care. When the GPU is probed an IOMMU group would be
created for the PE that contains it. We need to ensure that group is
removed before we add the PE to the compound group that's used to keep
the translations see by the PCIe and NVLink buses the same.
No functional changes. Probably.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-4-oohall@gmail.com
In pnv_ioda_setup_vf_PE() we register an iommu group for the VF PE
then call pnv_ioda_setup_bus_iommu_group() to add devices to that group.
However, this function is called before the VFs are scanned so there's
no devices to add.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-3-oohall@gmail.com
Re-work the control flow a bit so what's going on is a little clearer.
This also ensures the table_group is only initialised once in the P9
case. This shouldn't be a functional change since all the GPU PCI
devices should have the same table_group configuration, but it does
look strange.
Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200406030745.24595-2-oohall@gmail.com
Similar to the C code change, make the AMR restore conditional on
whether the register has changed.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429065654.1677541-7-npiggin@gmail.com
The AMR update is made conditional on AMR actually changing, which
should be the less common case on most workloads (though kernel page
faults on uaccess could be frequent, this doesn't significantly slow
down that case).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429065654.1677541-4-npiggin@gmail.com
Writing the AMR register is documented to require context
synchronizing operations before and after, for it to take effect as
expected. The KUAP restore at interrupt exit time deliberately avoids
the isync after the AMR update because it only needs to take effect
after the context synchronizing RFID that soon follows. Add a comment
for this.
The missing isync before the update doesn't have an obvious
justification, and seems it could theoretically allow a rogue user
access to leak past the AMR update. Add isyncs for these.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429065654.1677541-3-npiggin@gmail.com
Let's reduce the number of registers used in TLB miss handlers.
We have both r9 and r12 available for any temporary use.
r9 is enough, avoid using r12.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/7f330e971952abb2645fb9ca4310c0f527e84dcb.1590079969.git.christophe.leroy@csgroup.eu
Commit 1bc54c0311 ("powerpc: rework 4xx PTE access and TLB miss")
reworked 44x PTE access to avoid atomic pte updates, and
left 8xx, 40x and fsl booke with atomic pte updates.
Commit 6cfd8990e2 ("powerpc: rework FSL Book-E PTE access and TLB
miss") removed atomic pte updates on fsl booke.
It went away on 8xx with commit ddfc20a3b9 ("powerpc/8xx: Remove
PTE_ATOMIC_UPDATES").
40x is the last platform setting PTE_ATOMIC_UPDATES.
Rework PTE access and TLB miss to remove PTE_ATOMIC_UPDATES for 40x:
- Always handle DSI as a fault.
- Bail out of TLB miss handler when CONFIG_SWAP is set and
_PAGE_ACCESSED is not set.
- Bail out of ITLB miss handler when _PAGE_EXEC is not set.
- Only set WR bit when both _PAGE_RW and _PAGE_DIRTY are set.
- Remove _PAGE_HWWRITE
- Don't require PTE_ATOMIC_UPDATES anymore
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/99a0fcd337ef67088140d1647d75fea026a70413.1590079968.git.christophe.leroy@csgroup.eu
The latest Xilinx design tools called ISE and EDK has been released in
October 2013. New tool doesn't support any PPC405/PPC440 new designs.
These platforms are no longer supported and tested.
PowerPC 405/440 port is orphan from 2013 by
commit cdeb89943b ("MAINTAINERS: Fix incorrect status tag") and
commit 19624236cc ("MAINTAINERS: Update Grant's email address and maintainership")
that's why it is time to remove the support fot these platforms.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/8c593895e2cb57d232d85ce4d8c3a1aa7f0869cc.1590079968.git.christophe.leroy@csgroup.eu
The same complicated sequence for juggling EE, RI, soft mask, and
irq tracing is repeated 3 times, tidy these up into one function.
This differs qiute a bit between sub architectures, so this makes
the ppc32 port cleaner as well.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429062421.1675400-1-npiggin@gmail.com
The idea behind this prefetch was to kick off a page table walk before
returning from the fault, getting some pipelining advantage.
But this never showed up any noticable performance advantage, and in
fact with KUAP the prefetches are actually blocked and cause some
kind of micro-architectural fault. Removing this improves page fault
microbenchmark performance by about 9%.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[mpe: Keep the early return in update_mmu_cache()]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200504122907.49304-1-npiggin@gmail.com
The XIVE interrupt mode can be disabled with the "xive=off" kernel
parameter, in which case there is nothing to present to the user in the
associated /sys/kernel/debug/powerpc/xive file.
Fixes: 930914b7d5 ("powerpc/xive: Add a debugfs file to dump internal XIVE state")
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429075122.1216388-4-clg@kaod.org
Commit 1ca3dec2b2 ("powerpc/xive: Prevent page fault issues in the
machine crash handler") fixed an issue in the FW assisted dump of
machines using hash MMU and the XIVE interrupt mode under the POWER
hypervisor. It forced the mapping of the ESB page of interrupts being
mapped in the Linux IRQ number space to make sure the 'crash kexec'
sequence worked during such an event. But it didn't handle the
un-mapping.
This mapping is now blocking the removal of a passthrough IO adapter
under the POWER hypervisor because it expects the guest OS to have
cleared all page table entries related to the adapter. If some are
still present, the RTAS call which isolates the PCI slot returns error
9001 "valid outstanding translations".
Remove these mapping in the IRQ data cleanup routine.
Under KVM, this cleanup is not required because the ESB pages for the
adapter interrupts are un-mapped from the guest by the hypervisor in
the KVM XIVE native device. This is now redundant but it's harmless.
Fixes: 1ca3dec2b2 ("powerpc/xive: Prevent page fault issues in the machine crash handler")
Cc: stable@vger.kernel.org # v5.5+
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429075122.1216388-2-clg@kaod.org
The code patching code wants to get the value of a struct ppc_inst as
a u64 when the instruction is prefixed, so we can pass the u64 down to
__put_user_asm() and write it with a single store.
The optprobes code wants to load a struct ppc_inst as an immediate
into a register so it is useful to have it as a u64 to use the
existing helper function.
Currently this is a bit awkward because the value differs based on the
CPU endianness, so add a helper to do the conversion.
This fixes the usage in arch_prepare_optimized_kprobe() which was
previously incorrect on big endian.
Fixes: 650b55b707 ("powerpc: Add prefixed instructions to instruction data type")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Tested-by: Jordan Niethe <jniethe5@gmail.com>
Link: https://lore.kernel.org/r/20200526072630.2487363-1-mpe@ellerman.id.au
In a few places we want to calculate the address of the next
instruction. Previously that was simple, we just added 4 bytes, or if
using a u32 * we incremented that pointer by 1.
But prefixed instructions make it more complicated, we need to advance
by either 4 or 8 bytes depending on the actual instruction. We also
can't do pointer arithmetic using struct ppc_inst, because it is
always 8 bytes in size on 64-bit, even though we might only need to
advance by 4 bytes.
So add a ppc_inst_next() helper which calculates the location of the
next instruction, if the given instruction was located at the given
address. Note the instruction doesn't need to actually be at the
address in memory.
Although it would seem natural for the value to be passed by value,
that makes it too easy to write a loop that will read off the end of a
page, eg:
for (; src < end; src = ppc_inst_next(src, *src),
dest = ppc_inst_next(dest, *dest))
As noticed by Christophe and Jordan, if end is the exact end of a
page, and the next page is not mapped, this will fault, because *dest
will read 8 bytes, 4 bytes into the next page.
So value is passed by reference, so the helper can be careful to use
ppc_inst_read() on it.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Jordan Niethe <jniethe5@gmail.com>
Link: https://lore.kernel.org/r/20200522133318.1681406-1-mpe@ellerman.id.au
Merge our fixes branch from this cycle. It contains several important
fixes we need in next for testing purposes, and also some that will
conflict with upcoming changes.
Merge Christophe's large series to use huge pages for the linear
mapping on 8xx.
From his cover letter:
The main purpose of this big series is to:
- reorganise huge page handling to avoid using mm_slices.
- use huge pages to map kernel memory on the 8xx.
The 8xx supports 4 page sizes: 4k, 16k, 512k and 8M.
It uses 2 Level page tables, PGD having 1024 entries, each entry
covering 4M address space. Then each page table has 1024 entries.
At the time being, page sizes are managed in PGD entries, implying
the use of mm_slices as it can't mix several pages of the same size
in one page table.
The first purpose of this series is to reorganise things so that
standard page tables can also handle 512k pages. This is done by
adding a new _PAGE_HUGE flag which will be copied into the Level 1
entry in the TLB miss handler. That done, we have 2 types of pages:
- PGD entries to regular page tables handling 4k/16k and 512k pages
- PGD entries to hugepd tables handling 8M pages.
There is no need to mix 8M pages with other sizes, because a 8M page
will use more than what a single PGD covers.
Then comes the second purpose of this series. At the time being, the
8xx has implemented special handling in the TLB miss handlers in order
to transparently map kernel linear address space and the IMMR using
huge pages by building the TLB entries in assembly at the time of the
exception.
As mm_slices is only for user space pages, and also because it would
anyway not be convenient to slice kernel address space, it was not
possible to use huge pages for kernel address space. But after step
one of the series, it is now more flexible to use huge pages.
This series drop all assembly 'just in time' handling of huge pages
and use huge pages in page tables instead.
Once the above is done, then comes icing on the cake:
- Use huge pages for KASAN shadow mapping
- Allow pinned TLBs with strict kernel rwx
- Allow pinned TLBs with debug pagealloc
Then, last but not least, those modifications for the 8xx allows the
following improvement on book3s/32:
- Mapping KASAN shadow with BATs
- Allowing BATs with debug pagealloc
All this allows to considerably simplify TLB miss handlers and associated
initialisation. The overhead of reading page tables is negligible
compared to the reduction of the miss handlers.
While we were at touching pte_update(), some cleanup was done
there too.
Tested widely on 8xx and 832x. Boot tested on QEMU MAC99.
DEBUG_PAGEALLOC only manages RW data.
Text and RO data can still be mapped with BATs.
In order to map with BATs, also enforce data alignment. Set
by default to 256M which is a good compromise for keeping
enough BATs for also KASAN and IMMR.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/fd29c1718ee44d82115d0e835ced808eb4ccbf51.1589866984.git.christophe.leroy@csgroup.eu
DEBUG_PAGEALLOC only manages RW data.
Text and RO data can still be mapped with hugepages and pinned TLB.
In order to map with hugepages, also enforce a 512kB data alignment
minimum. That's a trade-off between size of speed, taking into
account that DEBUG_PAGEALLOC is a debug option. Anyway the alignment
is still tunable.
We also allow tuning of alignment for book3s to limit the complexity
of the test in Kconfig that will anyway disappear in the following
patches once DEBUG_PAGEALLOC is handled together with BATs.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c13256f2d356a316715da61fe089b3623ef217a5.1589866984.git.christophe.leroy@csgroup.eu
Pinned TLB are 8M. Now that there is no strict boundary anymore
between text and RO data, it is possible to use 8M pinned executable
TLB that covers both text and RO data.
When PIN_TLB_DATA or PIN_TLB_TEXT is selected, enforce 8M RW data
alignment and allow STRICT_KERNEL_RWX.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c535fc97bf0dd8693192e25feeed8088701e00c6.1589866984.git.christophe.leroy@csgroup.eu
Map linear memory space with 512k and 8M pages whenever
possible.
Three mappings are performed:
- One for kernel text
- One for RO data
- One for the rest
Separating the mappings is done to be able to update the
protection later when using STRICT_KERNEL_RWX.
The ITLB miss handler now need to also handle huge TLBs
unless kernel text in pinned.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c44f0ab5510474f25123d904cd1f4e5c6aa3c1ac.1589866984.git.christophe.leroy@csgroup.eu
Add a function to early map kernel memory using huge pages.
For 512k pages, just use standard page table and map in using 512k
pages.
For 8M pages, create a hugepd table and populate the two PGD
entries with it.
This function can only be used to create page tables at startup. Once
the regular SLAB allocation functions replace memblock functions,
this function cannot allocate new pages anymore. However it can still
update existing mappings with new protections.
hugepd_none() macro is moved into asm/hugetlb.h to be usable outside
of mm/hugetlbpage.c
early_pte_alloc_kernel() is made visible.
_PAGE_HUGE flag is now displayed by ptdump.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
[mpe: Change ptdump display to use "huge"]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/68325bcd3b6f93127f7810418a2352c3519066d6.1589866984.git.christophe.leroy@csgroup.eu
Now that linear and IMMR dedicated TLB handling is gone, kernel
boundary address comparison is similar in ITLB miss handler and
in DTLB miss handler.
Create a macro named compare_to_kernel_boundary.
When TASK_SIZE is strictly below 0x80000000 and PAGE_OFFSET is
above 0x80000000, it is enough to compare to 0x8000000, and this
can be done with a single instruction.
Using not. instruction, we get to use 'blt' conditional branch as
when doing a regular comparison:
0x00000000 <= addr <= 0x7fffffff ==>
0xffffffff >= NOT(addr) >= 0x80000000
The above test corresponds to a 'blt'
Otherwise, do a regular comparison using two instructions.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/6312575d06a8813105e6564a3b12e1d373aa1b2f.1589866984.git.christophe.leroy@csgroup.eu
Up to now, linear and IMMR mappings are managed via huge TLB entries
through specific code directly in TLB miss handlers. This implies
some patching of the TLB miss handlers at startup, and a lot of
dedicated code.
Remove all this specific dedicated code.
For now we are back to normal handling via standard 4k pages. In the
next patches, linear memory mapping and IMMR mapping will be managed
through huge pages.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/221b7e3ead80a5969629938c023f8cfe45fdd2fb.1589866984.git.christophe.leroy@csgroup.eu
At startup, map 32 Mbytes of memory through 4 pages of 8M,
and PIN them inconditionnaly. They need to be pinned because
KASAN is using page tables early and the TLBs might be
dynamically replaced otherwise.
Remove RSV4I flag after installing mappings unless
CONFIG_PIN_TLB_XXXX is selected.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/b27c5767d18053b59f7eefddc189fcc3acf7b9c2.1589866984.git.christophe.leroy@csgroup.eu
Only early debug requires IMMR to be mapped early.
No need to set it up and pin it in assembly. Map it
through page tables at udbg init when necessary.
If CONFIG_PIN_TLB_IMMR is selected, pin it once we
don't need the 32 Mb pinned RAM anymore.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/13c1e8539fdf363d3146f4884e5c3c76c6c308b5.1589866984.git.christophe.leroy@csgroup.eu
Pinned TLBs cannot be modified when the MMU is enabled.
Create a function to rewrite the pinned TLB entries with MMU off.
To set pinned TLB, we have to turn off MMU, disable pinning,
do a TLB flush (Either with tlbie and tlbia) then reprogam
the TLB entries, enable pinning and turn on MMU.
If using tlbie, it cleared entries in both instruction and data
TLB regardless whether pinning is disabled or not.
If using tlbia, it clears all entries of the TLB which has
disabled pinning.
To make it easy, just clear all entries in both TLBs, and
reprogram them.
The function takes two arguments, the top of the memory to
consider and whether data is RO under _sinittext.
When DEBUG_PAGEALLOC is set, the top is the end of kernel rodata.
Otherwise, that's the top of physical RAM.
Everything below _sinittext is set RX, over _sinittext that's RW.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c17806014bb1c06513ad1e1d510faea31984b177.1589866984.git.christophe.leroy@csgroup.eu
