Haren Myneni [Thu, 17 Jun 2021 20:35:22 +0000 (13:35 -0700)]
powerpc/pseries/vas: Add hcall wrappers for VAS handling
This patch adds the following hcall wrapper functions to allocate,
modify and deallocate VAS windows, and retrieve VAS capabilities.
H_ALLOCATE_VAS_WINDOW: Allocate VAS window
H_DEALLOCATE_VAS_WINDOW: Close VAS window
H_MODIFY_VAS_WINDOW: Setup window before using
H_QUERY_VAS_CAPABILITIES: Get VAS capabilities
Haren Myneni [Thu, 17 Jun 2021 20:34:43 +0000 (13:34 -0700)]
powerpc/vas: Define QoS credit flag to allocate window
PowerVM introduces two different type of credits: Default and Quality
of service (QoS).
The total number of default credits available on each LPAR depends
on CPU resources configured. But these credits can be shared or
over-committed across LPARs in shared mode which can result in
paste command failure (RMA_busy). To avoid NX HW contention, the
hypervisor ntroduces QoS credit type which makes sure guaranteed
access to NX esources. The system admins can assign QoS credits
or each LPAR via HMC.
Default credit type is used to allocate a VAS window by default as
on PowerVM implementation. But the process can pass
VAS_TX_WIN_FLAG_QOS_CREDIT flag with VAS_TX_WIN_OPEN ioctl to open
QoS type window.
Haren Myneni [Thu, 17 Jun 2021 20:33:28 +0000 (13:33 -0700)]
powerpc/vas: Define and use common vas_window struct
Many elements in vas_struct are used on PowerNV and PowerVM
platforms. vas_window is used for both TX and RX windows on
PowerNV and for TX windows on PowerVM. So some elements are
specific to these platforms.
So this patch defines common vas_window and platform
specific window structs (pnv_vas_window on PowerNV). Also adds
the corresponding changes in PowerNV vas code.
Haren Myneni [Thu, 17 Jun 2021 20:32:38 +0000 (13:32 -0700)]
powerpc/vas: Move update_csb/dump_crb to common book3s platform
If a coprocessor encounters an error translating an address, the
VAS will cause an interrupt in the host. The kernel processes
the fault by updating CSB. This functionality is same for both
powerNV and pseries. So this patch moves these functions to
common vas-api.c and the actual functionality is not changed.
Haren Myneni [Thu, 17 Jun 2021 20:31:43 +0000 (13:31 -0700)]
powerpc/vas: Create take/drop pid and mm reference functions
Take pid and mm references when each window opens and drops during
close. This functionality is needed for powerNV and pseries. So
this patch defines the existing code as functions in common book3s
platform vas-api.c
powerNV and pseries drivers register / unregister to the corresponding
platform specific VAS separately. Then these VAS functions call the
common API with the specific window operations. So rename powerNV VAS
API register/unregister functions.
Haren Myneni [Thu, 17 Jun 2021 20:29:48 +0000 (13:29 -0700)]
powerpc/vas: Move VAS API to book3s common platform
The pseries platform will share vas and nx code and interfaces
with the PowerNV platform, so create the
arch/powerpc/platforms/book3s/ directory and move VAS API code
there. Functionality is not changed.
Haren Myneni [Thu, 17 Jun 2021 20:29:05 +0000 (13:29 -0700)]
powerpc/powernv/vas: Release reference to tgid during window close
The kernel handles the NX fault by updating CSB or sending
signal to process. In multithread applications, children can
open VAS windows and can exit without closing them. But the
parent can continue to send NX requests with these windows. To
prevent pid reuse, reference will be taken on pid and tgid
when the window is opened and release them during window close.
The current code is not releasing the tgid reference which can
cause pid leak and this patch fixes the issue.
Fixes: db1c08a740635 ("powerpc/vas: Take reference to PID and mm for user space windows") Cc: stable@vger.kernel.org # 5.8+ Reported-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Haren Myneni <haren@linux.ibm.com> Reviewed-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/6020fc4d444864fe20f7dcdc5edfe53e67480a1c.camel@linux.ibm.com
Update _tlbiel_pid() such that we can avoid build errors like below when
using this function in other places.
arch/powerpc/mm/book3s64/radix_tlb.c: In function ‘__radix__flush_tlb_range_psize’:
arch/powerpc/mm/book3s64/radix_tlb.c:114:2: warning: ‘asm’ operand 3 probably does not match constraints
114 | asm volatile(PPC_TLBIEL(%0, %4, %3, %2, %1)
| ^~~
arch/powerpc/mm/book3s64/radix_tlb.c:114:2: error: impossible constraint in ‘asm’
make[4]: *** [scripts/Makefile.build:271: arch/powerpc/mm/book3s64/radix_tlb.o] Error 1
m
With this fix, we can also drop the __always_inline in __radix_flush_tlb_range_psize
which was added by commit e12d6d7d46a6 ("powerpc/mm/radix: mark __radix__flush_tlb_range_psize() as __always_inline")
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu> Acked-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20210610083639.387365-1-aneesh.kumar@linux.ibm.com
Michael Ellerman [Thu, 10 Jun 2021 07:29:49 +0000 (17:29 +1000)]
powerpc/signal64: Don't read sigaction arguments back from user memory
When delivering a signal to a sigaction style handler (SA_SIGINFO), we
pass pointers to the siginfo and ucontext via r4 and r5.
Currently we populate the values in those registers by reading the
pointers out of the sigframe in user memory, even though the values in
user memory were written by the kernel just prior:
unsafe_put_user(&frame->info, &frame->pinfo, badframe_block);
unsafe_put_user(&frame->uc, &frame->puc, badframe_block);
...
if (ksig->ka.sa.sa_flags & SA_SIGINFO) {
err |= get_user(regs->gpr[4], (unsigned long __user *)&frame->pinfo);
err |= get_user(regs->gpr[5], (unsigned long __user *)&frame->puc);
ie. we write &frame->info into frame->pinfo, and then read frame->pinfo
back into r4, and similarly for &frame->uc.
The code has always been like this, since linux-fullhistory commit d4f2d95eca2c ("Forward port of 2.4 ppc64 signal changes.").
There's no reason for us to read the values back from user memory,
rather than just setting the value in the gpr[4/5] directly. In fact
reading the value back from user memory opens up the possibility of
another user thread changing the values before we read them back.
Although any process doing that would be racing against the kernel
delivering the signal, and would risk corrupting the stack, so that
would be a userspace bug.
Note that this is 64-bit only code, so there's no subtlety with the size
of pointers differing between kernel and user. Also the frame variable
is not modified to point elsewhere during the function.
In the past reading the values back from user memory was not costly, but
now that we have KUAP on some CPUs it is, so we'd rather avoid it for
that reason too.
So change the code to just set the values directly, using the same
values we have written to the sigframe previously in the function.
Note also that this matches what our 32-bit signal code does.
Using a version of will-it-scale's signal1_threads that sets SA_SIGINFO,
this results in a ~4% increase in signals per second on a Power9, from
229,777 to 239,766.
Daniel Axtens [Mon, 14 Jun 2021 12:09:07 +0000 (22:09 +1000)]
powerpc: make stack walking KASAN-safe
Make our stack-walking code KASAN-safe by using __no_sanitize_address.
Generic code, arm64, s390 and x86 all make accesses unchecked for similar
sorts of reasons: when unwinding a stack, we might touch memory that KASAN
has marked as being out-of-bounds. In ppc64 KASAN development, I hit this
sometimes when checking for an exception frame - because we're checking
an arbitrary offset into the stack frame.
See commit 20955746320e ("s390/kasan: avoid false positives during stack
unwind"), commit bcaf669b4bdb ("arm64: disable kasan when accessing
frame->fp in unwind_frame"), commit 91e08ab0c851 ("x86/dumpstack:
Prevent KASAN false positive warnings") and commit 6e22c8366416
("tracing, kasan: Silence Kasan warning in check_stack of stack_tracer").
Athira Rajeev [Tue, 25 May 2021 13:51:43 +0000 (09:51 -0400)]
selftests/powerpc: EBB selftest for MMCR0 control for PMU SPRs in ISA v3.1
With the MMCR0 control bit (PMCCEXT) in ISA v3.1, read access to
group B registers is restricted when MMCR0 PMCC=0b00. In other
platforms (like power9), the older behaviour works where group B
PMU SPRs are readable.
Patch creates a selftest which verifies that the test takes a
SIGILL when attempting to read PMU registers via helper function
"dump_ebb_state" for ISA v3.1.
Athira Rajeev [Tue, 25 May 2021 13:51:42 +0000 (09:51 -0400)]
selftests/powerpc: Fix "no_handler" EBB selftest
The "no_handler_test" in ebb selftests attempts to read the PMU
registers twice via helper function "dump_ebb_state". First dump is
just before closing of event and the second invocation is done after
closing of the event. The original intention of second
dump_ebb_state was to dump the state of registers at the end of
the test when the counters are frozen. But this will be achieved
with the first call itself since sample period is set to low value
and PMU will be frozen by then. Hence patch removes the
dump which was done before closing of the event.
Reported-by: Shirisha Ganta <shirisha.ganta1@ibm.com> Signed-off-by: Athira Rajeev <atrajeev@linux.vnet.ibm.com> Tested-by: Nageswara R Sastry <rnsastry@linux.ibm.com <mailto:rnsastry@linux.ibm.com>> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/1621950703-1532-2-git-send-email-atrajeev@linux.vnet.ibm.com
Christophe Leroy [Mon, 7 Jun 2021 10:56:04 +0000 (10:56 +0000)]
powerpc: Define empty_zero_page[] in C
At the time being, empty_zero_page[] is defined in each
platform head.S.
Define it in mm/mem.c instead, and put it in BSS section instead
of the DATA section. Commit 5227cfa71f9e ("arm64: mm: place
empty_zero_page in bss") explains why it is interesting to have
it in BSS.
Christophe Leroy [Thu, 3 Jun 2021 09:29:06 +0000 (09:29 +0000)]
powerpc/nohash: Remove DEBUG_CLAMP_LAST_CONTEXT
DEBUG_CLAMP_LAST_CONTEXT was there in the old days to reduce
number of contexts in order to ease debugging implementation
of context switching, but that's been quite stable during
years now.
Christophe Leroy [Thu, 3 Jun 2021 09:13:54 +0000 (09:13 +0000)]
powerpc/kuap: Force inlining of all first level KUAP helpers.
All KUAP helpers defined in asm/kup.h are single line functions
that should be inlined. But on book3s/32 build, we get many
instances of <prevent_write_to_user.constprop.0>.
Christophe Leroy [Thu, 3 Jun 2021 08:41:45 +0000 (08:41 +0000)]
powerpc/32s: Activate KUAP and KUEP by default
Now that KUAP and KUEP have been significantly optimised and can be
disabled at boot time using 'nosmap' and 'nosmep' kernel parameters,
them can be active by default like in other powerpc platforms.
It is still possible to disable them completely in the configuration.
On book3s/32, KUAP is provided by toggling Ks bit in segment registers.
One segment register addresses 256M of virtual memory.
At the time being, KUAP implements a complex logic to apply the
unlock/lock on the exact number of segments covering the user range
to access, with saving the boundaries of the range of segments in
a member of thread struct.
But most if not all user accesses are within a single segment.
Rework KUAP with a different approach:
- Open only one segment, the one corresponding to the starting
address of the range to be accessed.
- If a second segment is involved, it will generate a page fault. The
segment will then be open by the page fault handler.
The kuap member of thread struct will now contain:
- The start address of the current on going user access, that will be
used to know which segment to lock at the end of the user access.
- ~0 when no user access is open
- ~1 when additionnal segments are opened by a page fault.
Then, at lock time
- When only one segment is open, close it.
- When several segments are open, close all user segments.
Almost 100% of the time, only one segment will be involved.
In interrupts, inline the function that unlock/lock all segments,
because not inlining them implies a lot of register save/restore.
With the patch, writing value 128 in userspace in perf_copy_attr() is
done with 16 instructions:
Before the patch it was 118 instructions. In reality only 42 are
executed in most cases, but GCC is not able to see that a properly
aligned user access cannot involve more than one segment.
Christophe Leroy [Thu, 3 Jun 2021 08:41:43 +0000 (08:41 +0000)]
powerpc/32s: Allow disabling KUAP at boot time
PPC64 uses MMU features to enable/disable KUAP at boot time.
But feature fixups are applied way too early on PPC32.
Now that all KUAP related actions are in C following the
conversion of KUAP initial setup and context switch in C,
static branches can be used to enable/disable KUAP.
Christophe Leroy [Thu, 3 Jun 2021 08:41:42 +0000 (08:41 +0000)]
powerpc/32s: Allow disabling KUEP at boot time
PPC64 uses MMU features to enable/disable KUEP at boot time.
But feature fixups are applied way too early on PPC32.
Now that all KUEP related actions are in C following the
conversion of KUEP initial setup and context switch in C,
static branches can be used to enable/disable KUEP.
Christophe Leroy [Thu, 3 Jun 2021 08:41:39 +0000 (08:41 +0000)]
powerpc/32s: Convert switch_mmu_context() to C
switch_mmu_context() does things that can easily be done in C.
For updating user segments, we have update_user_segments().
As mentionned in commit b5efec00b671 ("powerpc/32s: Move KUEP
locking/unlocking in C"), update_user_segments() has the loop
unrolled which is a significant performance gain.
Christophe Leroy [Thu, 3 Jun 2021 08:41:37 +0000 (08:41 +0000)]
powerpc/32s: Refactor update of user segment registers
KUEP implements the update of user segment registers.
Move it into mmu-hash.h in order to use it from other places.
And inline kuep_lock() and kuep_unlock(). Inlining kuep_lock() is
important for system_call_exception(), otherwise system_call_exception()
has to save into stack the system call parameters that are used just
after, and doing that takes more instructions than kuep_lock() itself.
Christophe Leroy [Thu, 20 May 2021 13:50:47 +0000 (13:50 +0000)]
powerpc/optprobes: Minimise casts
nip is already an unsigned long, no cast needed.
op_callback_addr and emulate_step_addr are kprobe_opcode_t *.
There value is obtained with ppc_kallsyms_lookup_name() which
returns 'unsigned long', and there values are used create_branch()
which expects 'unsigned long'. So change them to 'unsigned long'
to avoid casting them back and forth.
can_optimize() used p->addr several times as 'unsigned long'.
Use a local 'unsigned long' variable and avoid casting multiple times.
Christophe Leroy [Thu, 20 May 2021 13:50:42 +0000 (13:50 +0000)]
powerpc: Do not dereference code as 'struct ppc_inst' (uprobe, code-patching, feature-fixups)
'struct ppc_inst' is an internal structure to represent an instruction,
it is not directly the representation of that instruction in text code.
It is not meant to map and dereference code.
Dereferencing code directly through 'struct ppc_inst' has two main issues:
- On powerpc, structs are expected to be 8 bytes aligned while code is
spread every 4 byte.
- Should a non prefixed instruction lie at the end of the page and the
following page not be mapped, it would generate a page fault.
In-memory code must be accessed with ppc_inst_read().
Christophe Leroy [Thu, 20 May 2021 13:50:41 +0000 (13:50 +0000)]
powerpc/inst: Avoid pointer dereferencing in ppc_inst_equal()
Avoid casting/dereferencing ppc_inst() as u64* , check each member
of the struct when relevant.
And remove the 0xff initialisation of the suffix for non
prefixed instruction. An instruction with 0xff as a suffix
might be invalid, but still is a prefixed instruction and
has to be considered as this.
Christophe Leroy [Thu, 20 May 2021 10:23:10 +0000 (10:23 +0000)]
powerpc/traps: Start using PPC_RAW_xx() macros
Start using PPC_RAW_xx() macros where relevant.
PPC_INST_SYNC is used to both represent the 'sync' instruction and
the family of synchronisation instructions. Keep it for the later,
maybe we'll change the name in the future to avoid confusion.
Christophe Leroy [Thu, 20 May 2021 10:23:00 +0000 (10:23 +0000)]
powerpc: Rework PPC_RAW_xxx() macros for prefixed instructions
At the time being, we have PPC_RAW_PLXVP() and PPC_RAW_PSTXVP() which
provide a 64 bits value, and then it gets split by open coding to
format it into a 'struct ppc_inst' instruction.
Instead, define a PPC_RAW_xxx_P() and a PPC_RAW_xxx_S() to be used
as is.
Christophe Leroy [Fri, 14 May 2021 13:14:53 +0000 (13:14 +0000)]
powerpc: Don't handle ALTIVEC/SPE in ASM in _switch(). Do it in C.
_switch() saves and restores ALTIVEC and SPE status.
For altivec this is redundant with what __switch_to() does with
save_sprs() and restore_sprs() and giveup_all() before
calling _switch().
Add support for SPI in save_sprs() and restore_sprs() and
remove things from _switch().
The sum with 0 is useless, should have been skipped.
And there is even one completely unused instance of csum_add().
In file included from ./include/net/checksum.h:22,
from ./include/linux/skbuff.h:28,
from ./include/linux/icmp.h:16,
from net/ipv6/ip6_tunnel.c:23:
./arch/powerpc/include/asm/checksum.h: In function '__ip6_tnl_rcv':
./arch/powerpc/include/asm/checksum.h:94:22: warning: inlining failed in call to 'csum_add': call is unlikely and code size would grow [-Winline]
94 | static inline __wsum csum_add(__wsum csum, __wsum addend)
| ^~~~~~~~
./arch/powerpc/include/asm/checksum.h:172:31: note: called from here
172 | sum = csum_add(sum, (__force __wsum)*(const u32 *)buff);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
./arch/powerpc/include/asm/checksum.h:94:22: warning: inlining failed in call to 'csum_add': call is unlikely and code size would grow [-Winline]
94 | static inline __wsum csum_add(__wsum csum, __wsum addend)
| ^~~~~~~~
./arch/powerpc/include/asm/checksum.h:177:31: note: called from here
177 | sum = csum_add(sum, (__force __wsum)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
178 | *(const u32 *)(buff + 4));
| ~~~~~~~~~~~~~~~~~~~~~~~~~
./arch/powerpc/include/asm/checksum.h:94:22: warning: inlining failed in call to 'csum_add': call is unlikely and code size would grow [-Winline]
94 | static inline __wsum csum_add(__wsum csum, __wsum addend)
| ^~~~~~~~
./arch/powerpc/include/asm/checksum.h:183:31: note: called from here
183 | sum = csum_add(sum, (__force __wsum)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
184 | *(const u32 *)(buff + 8));
| ~~~~~~~~~~~~~~~~~~~~~~~~~
./arch/powerpc/include/asm/checksum.h:94:22: warning: inlining failed in call to 'csum_add': call is unlikely and code size would grow [-Winline]
94 | static inline __wsum csum_add(__wsum csum, __wsum addend)
| ^~~~~~~~
./arch/powerpc/include/asm/checksum.h:186:31: note: called from here
186 | sum = csum_add(sum, (__force __wsum)
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
187 | *(const u16 *)(buff + 12));
| ~~~~~~~~~~~~~~~~~~~~~~~~~~
Finn Thain [Fri, 11 Jun 2021 07:58:27 +0000 (17:58 +1000)]
powerpc/tau: Remove superfluous parameter in alloc_workqueue() call
This avoids an (optional) compiler warning:
arch/powerpc/kernel/tau_6xx.c: In function 'TAU_init':
arch/powerpc/kernel/tau_6xx.c:204:30: error: too many arguments for format [-Werror=format-extra-args]
tau_workq = alloc_workqueue("tau", WQ_UNBOUND, 1, 0);
Michael Ellerman [Mon, 14 Jun 2021 13:14:40 +0000 (23:14 +1000)]
powerpc: Fix initrd corruption with relative jump labels
Commit b0b3b2c78ec0 ("powerpc: Switch to relative jump labels") switched
us to using relative jump labels. That involves changing the code,
target and key members in struct jump_entry to be relative to the
address of the jump_entry, rather than absolute addresses.
We have two static inlines that create a struct jump_entry,
arch_static_branch() and arch_static_branch_jump(), as well as an asm
macro ARCH_STATIC_BRANCH, which is used by the pseries-only hypervisor
tracing code.
Unfortunately we missed updating the key to be a relative reference in
ARCH_STATIC_BRANCH.
That causes a pseries kernel to have a handful of jump_entry structs
with bad key values. Instead of being a relative reference they instead
hold the full address of the key.
However the code doesn't expect that, it still adds the key value to the
address of the jump_entry (see jump_entry_key()) expecting to get a
pointer to a key somewhere in kernel data.
The table of jump_entry structs sits in rodata, which comes after the
kernel text. In a typical build this will be somewhere around 15MB. The
address of the key will be somewhere in data, typically around 20MB.
Adding the two values together gets us a pointer somewhere around 45MB.
We then call static_key_set_entries() with that bad pointer and modify
some members of the struct static_key we think we are pointing at.
A pseries kernel is typically ~30MB in size, so writing to ~45MB won't
corrupt the kernel itself. However if we're booting with an initrd,
depending on the size and exact location of the initrd, we can corrupt
the initrd. Depending on how exactly we corrupt the initrd it can either
cause the system to not boot, or just corrupt one of the files in the
initrd.
The fix is simply to make the key value relative to the jump_entry
struct in the ARCH_STATIC_BRANCH macro.
Fixes: b0b3b2c78ec0 ("powerpc: Switch to relative jump labels") Reported-by: Anastasia Kovaleva <a.kovaleva@yadro.com> Reported-by: Roman Bolshakov <r.bolshakov@yadro.com> Reported-by: Greg Kurz <groug@kaod.org> Reported-by: Daniel Axtens <dja@axtens.net> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Tested-by: Daniel Axtens <dja@axtens.net> Tested-by: Greg Kurz <groug@kaod.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20210614131440.312360-1-mpe@ellerman.id.au
Christophe Leroy [Fri, 7 May 2021 14:01:09 +0000 (14:01 +0000)]
powerpc/perf: Simplify Makefile
arch/powerpc/Kbuild decend into arch/powerpc/perf/ only when
CONFIG_PERF_EVENTS is selected, so there is not need to take
CONFIG_PERF_EVENTS into account in arch/powerpc/perf/Makefile.
Baokun Li [Tue, 1 Jun 2021 08:53:19 +0000 (16:53 +0800)]
powerpc/spider-pci: Remove set but not used variable 'val'
Fixes gcc '-Wunused-but-set-variable' warning:
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
arch/powerpc/platforms/cell/spider-pci.c: In function 'spiderpci_io_flush':
arch/powerpc/platforms/cell/spider-pci.c:28:6: warning:
variable ‘val’ set but not used [-Wunused-but-set-variable]
Baokun Li [Tue, 1 Jun 2021 08:51:27 +0000 (16:51 +0800)]
powerpc/spufs: Remove set but not used variable 'dummy'
Fixes gcc '-Wunused-but-set-variable' warning:
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
# WARNING: Fixes tag on line 3 doesn't match correct format
arch/powerpc/platforms/cell/spufs/switch.c: In function 'check_ppu_mb_stat':
arch/powerpc/platforms/cell/spufs/switch.c:1660:6: warning:
variable ‘dummy’ set but not used [-Wunused-but-set-variable]
arch/powerpc/platforms/cell/spufs/switch.c: In function 'check_ppuint_mb_stat':
arch/powerpc/platforms/cell/spufs/switch.c:1675:6: warning:
variable ‘dummy’ set but not used [-Wunused-but-set-variable]
Michael Ellerman [Tue, 8 Jun 2021 13:46:05 +0000 (23:46 +1000)]
powerpc/signal64: Copy siginfo before changing regs->nip
In commit 96d7a4e06fab ("powerpc/signal64: Rewrite handle_rt_signal64()
to minimise uaccess switches") the 64-bit signal code was rearranged to
use user_write_access_begin/end().
As part of that change the call to copy_siginfo_to_user() was moved
later in the function, so that it could be done after the
user_write_access_end().
In particular it was moved after we modify regs->nip to point to the
signal trampoline. That means if copy_siginfo_to_user() fails we exit
handle_rt_signal64() with an error but with regs->nip modified, whereas
previously we would not modify regs->nip until the copy succeeded.
Returning an error from signal delivery but with regs->nip updated
leaves the process in a sort of half-delivered state. We do immediately
force a SEGV in signal_setup_done(), called from do_signal(), so the
process should never run in the half-delivered state.
However that SEGV is not delivered until we've gone around to
do_notify_resume() again, so it's possible some tracing could observe
the half-delivered state.
There are other cases where we fail signal delivery with regs partly
updated, eg. the write to newsp and SA_SIGINFO, but the latter at least
is very unlikely to fail as it reads back from the frame we just wrote
to.
Looking at other arches they seem to be more careful about leaving regs
unchanged until the copy operations have succeeded, and in general that
seems like good hygenie.
So although the current behaviour is not cleary buggy, it's also not
clearly correct. So move the call to copy_siginfo_to_user() up prior to
the modification of regs->nip, which is closer to the old behaviour, and
easier to reason about.
Implement support for hash guests under hash host. This has to save and
restore the host SLB, and ensure that the MMU is off while switching
into the guest SLB.
POWER9 and later CPUs now always go via the P9 path. The "fast" guest
mode is now renamed to the P9 mode, which is consistent with its
functionality and the rest of the naming.
Nicholas Piggin [Fri, 28 May 2021 09:07:50 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV P9: implement hash guest support
Implement hash guest support. Guest entry/exit has to restore and
save/clear the SLB, plus several other bits to accommodate hash guests
in the P9 path. Radix host, hash guest support is removed from the P7/8
path.
The HPT hcalls and faults are not handled in real mode, which is a
performance regression. A worst-case fork/exit microbenchmark takes 3x
longer after this patch. kbuild benchmark performance is in the noise,
but the slowdown is likely to be noticed somewhere.
For now, accept this penalty for the benefit of simplifying the P7/8
paths and unifying P9 hash with the new code, because hash is a less
important configuration than radix on processors that support it. Hash
will benefit from future optimisations to this path, including possibly
a faster path to handle such hcalls and interrupts without doing a full
exit.
Nicholas Piggin [Fri, 28 May 2021 09:07:49 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV P9: Reflect userspace hcalls to hash guests to support PR KVM
The reflection of sc 1 interrupts from guest PR=1 to the guest kernel is
required to support a hash guest running PR KVM where its guest is
making hcalls with sc 1.
In preparation for hash guest support, add this hcall reflection to the
P9 path. The P7/8 path does this in its realmode hcall handler
(sc_1_fast_return).
Nicholas Piggin [Fri, 28 May 2021 09:07:48 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV: add virtual mode handlers for HPT hcalls and page faults
In order to support hash guests in the P9 path (which does not do real
mode hcalls or page fault handling), these real-mode hash specific
interrupts need to be implemented in virt mode.
Commit f3c18e9342a44 ("KVM: PPC: Book3S HV: Use XICS hypercalls when
running as a nested hypervisor") added nested HV tests in XICS
hypercalls, but not all are required.
* icp_eoi is only called by kvmppc_deliver_irq_passthru which is only
called by kvmppc_check_passthru which is only caled by
kvmppc_read_one_intr.
* kvmppc_read_one_intr is only called by kvmppc_read_intr which is only
called by the L0 HV rmhandlers code.
* kvmhv_rm_send_ipi is called by:
- kvmhv_interrupt_vcore which is only called by kvmhv_commence_exit
which is only called by the L0 HV rmhandlers code.
- icp_send_hcore_msg which is only called by icp_rm_set_vcpu_irq.
- icp_rm_set_vcpu_irq which is only called by icp_rm_try_update
- icp_rm_set_vcpu_irq is not nested HV safe because it writes to
LPCR directly without a kvmhv_on_pseries test. Nested handlers
should not in general be using the rm handlers.
The important test seems to be in kvmppc_ipi_thread, which sends the
virt-mode H_IPI handler kick to use smp_call_function rather than
msgsnd.
Nicholas Piggin [Fri, 28 May 2021 09:07:44 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV: Remove virt mode checks from real mode handlers
Now that the P7/8 path no longer supports radix, real-mode handlers
do not need to deal with being called in virt mode.
This change effectively reverts commit acde25726bc6 ("KVM: PPC: Book3S
HV: Add radix checks in real-mode hypercall handlers").
It removes a few more real-mode tests in rm hcall handlers, which
allows the indirect ops for the xive module to be removed from the
built-in xics rm handlers.
kvmppc_h_random is renamed to kvmppc_rm_h_random to be a bit more
descriptive and consistent with other rm handlers.
Nicholas Piggin [Fri, 28 May 2021 09:07:42 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV: Remove support for dependent threads mode on P9
Dependent-threads mode is the normal KVM mode for pre-POWER9 SMT
processors, where all threads in a core (or subcore) would run the same
partition at the same time, or they would run the host.
This design was mandated by MMU state that is shared between threads in
a processor, so the synchronisation point is in hypervisor real-mode
that has essentially no shared state, so it's safe for multiple threads
to gather and switch to the correct mode.
It is implemented by having the host unplug all secondary threads and
always run in SMT1 mode, and host QEMU threads essentially represent
virtual cores that wake these secondary threads out of unplug when the
ioctl is called to run the guest. This happens via a side-path that is
mostly invisible to the rest of the Linux host and the secondary threads
still appear to be unplugged.
POWER9 / ISA v3.0 has a more flexible MMU design that is independent
per-thread and allows a much simpler KVM implementation. Before the new
"P9 fast path" was added that began to take advantage of this, POWER9
support was implemented in the existing path which has support to run
in the dependent threads mode. So it was not much work to add support to
run POWER9 in this dependent threads mode.
The mode is not required by the POWER9 MMU (although "mixed-mode" hash /
radix MMU limitations of early processors were worked around using this
mode). But it is one way to run SMT guests without running different
guests or guest and host on different threads of the same core, so it
could avoid or reduce some SMT attack surfaces without turning off SMT
entirely.
This security feature has some real, if indeterminate, value. However
the old path is lagging in features (nested HV), and with this series
the new P9 path adds remaining missing features (radix prefetch bug
and hash support, in later patches), so POWER9 dependent threads mode
support would be the only remaining reason to keep that code in and keep
supporting POWER9/POWER10 in the old path. So here we make the call to
drop this feature.
Remove dependent threads mode support for POWER9 and above processors.
Systems can still achieve this security by disabling SMT entirely, but
that would generally come at a larger performance cost for guests.
Rather than partition the guest PID space + flush a rogue guest PID to
work around this problem, instead fix it by always disabling the MMU when
switching in or out of guest MMU context in HV mode.
This may be a bit less efficient, but it is a lot less complicated and
allows the P9 path to trivally implement the workaround too. Newer CPUs
are not subject to this issue.
Nicholas Piggin [Fri, 28 May 2021 09:07:40 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV P9: Switch to guest MMU context as late as possible
Move MMU context switch as late as reasonably possible to minimise code
running with guest context switched in. This becomes more important when
this code may run in real-mode, with later changes.
Move WARN_ON as early as possible so program check interrupts are less
likely to tangle everything up.
Nicholas Piggin [Fri, 28 May 2021 09:07:39 +0000 (19:07 +1000)]
KVM: PPC: Book3S HV P9: Add helpers for OS SPR handling
This is a first step to wrapping supervisor and user SPR saving and
loading up into helpers, which will then be called independently in
bare metal and nested HV cases in order to optimise SPR access.
projects / users / jedix / linux-maple.git / log