Fix a cut'n'paste error in a warning message. This should be
'cpu-idle-state-residency-ns' to match the property searched in the
previous 'of_property_read_u32_array()'
Fixes: 9c7b185ab2 ("powernv/cpuidle: Parse dt idle properties into global structure")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200502115949.139000-1-christophe.jaillet@wanadoo.fr
This merges the lockless page table walk rework series from Aneesh.
Because it touches powerpc KVM code we are sharing it with the kvm-ppc
tree in our topic/ppc-kvm branch.
This is the cover letter from Aneesh:
Avoid IPI while updating page table entries.
Problem Summary:
Slow termination of KVM guest with large guest RAM config due to a
large number of IPIs that were caused by clearing level 1 PTE
entries (THP) entries. This is shown in the stack trace below.
- qemu-system-ppc [kernel.vmlinux] [k] smp_call_function_many
- smp_call_function_many
- 36.09% smp_call_function_many
serialize_against_pte_lookup
radix__pmdp_huge_get_and_clear
zap_huge_pmd
unmap_page_range
unmap_vmas
unmap_region
__do_munmap
__vm_munmap
sys_munmap
system_call
__munmap
qemu_ram_munmap
qemu_anon_ram_free
reclaim_ramblock
call_rcu_thread
qemu_thread_start
start_thread
__clone
Why we need to do IPI when clearing PMD entries:
This was added as part of commit: 13bd817bb8 ("powerpc/thp: Serialize pmd clear against a linux page table walk")
serialize_against_pte_lookup makes sure that all parallel lockless
page table walk completes before we convert a PMD pte entry to regular
pmd entry. We end up doing that conversion in the below scenarios
1) __split_huge_zero_page_pmd
2) do_huge_pmd_wp_page_fallback
3) MADV_DONTNEED running parallel to page faults.
local_irq_disable and lockless page table walk:
The lockless page table walk work with the assumption that we can
dereference the page table contents without holding a lock. For this
to work, we need to make sure we read the page table contents
atomically and page table pages are not going to be freed/released
while we are walking the table pages. We can achieve by using a rcu
based freeing for page table pages or if the architecture implements
broadcast tlbie, we can block the IPI as we walk the page table pages.
To support both the above framework, lockless page table walk is done
with irq disabled instead of rcu_read_lock()
We do have two interface for lockless page table walk, gup fast and
__find_linux_pte. This patch series makes __find_linux_pte table walk
safe against the conversion of PMD PTE to regular PMD.
gup fast:
gup fast is already safe against THP split because kernel now
differentiate between a pmd split and a compound page split. gup fast
can run parallel to a pmd split and we prevent a parallel gup fast to
a hugepage split, by freezing the page refcount and failing the
speculative page ref increment.
Similar to how gup is safe against parallel pmd split, this patch
series updates the __find_linux_pte callers to be safe against a
parallel pmd split. We do that by enforcing the following rules.
1) Don't reload the pte value, because that can be updated in
parallel.
2) Code should be able to work with a stale PTE value and not the
recent one. ie, the pte value that we are looking at may not be the
latest value in the page table.
3) Before looking at pte value check for _PAGE_PTE bit. We now do this
as part of pte_present() check.
Performance:
This speeds up Qemu guest RAM del/unplug time as below
128 core, 496GB guest:
Without patch:
munmap start: timer = 13162 ms, PID=7684
munmap finish: timer = 95312 ms, PID=7684 - delta = 82150 ms
With patch (upto removing IPI)
munmap start: timer = 196449 ms, PID=6681
munmap finish: timer = 196488 ms, PID=6681 - delta = 39ms
With patch (with adding the tlb invalidate in pmdp_huge_get_and_clear_full)
munmap start: timer = 196345 ms, PID=6879
munmap finish: timer = 196714 ms, PID=6879 - delta = 369ms
Link: https://lore.kernel.org/r/20200505071729.54912-1-aneesh.kumar@linux.ibm.com
MADV_DONTNEED holds mmap_sem in read mode and that implies a
parallel page fault is possible and the kernel can end up with a level 1 PTE
entry (THP entry) converted to a level 0 PTE entry without flushing
the THP TLB entry.
Most architectures including POWER have issues with kernel instantiating a level
0 PTE entry while holding level 1 TLB entries.
The code sequence I am looking at is
down_read(mmap_sem) down_read(mmap_sem)
zap_pmd_range()
zap_huge_pmd()
pmd lock held
pmd_cleared
table details added to mmu_gather
pmd_unlock()
insert a level 0 PTE entry()
tlb_finish_mmu().
Fix this by forcing a tlb flush before releasing pmd lock if this is
not a fullmm invalidate. We can safely skip this invalidate for
task exit case (fullmm invalidate) because in that case we are sure
there can be no parallel fault handlers.
This do change the Qemu guest RAM del/unplug time as below
128 core, 496GB guest:
Without patch:
munmap start: timer = 196449 ms, PID=6681
munmap finish: timer = 196488 ms, PID=6681 - delta = 39ms
With patch:
munmap start: timer = 196345 ms, PID=6879
munmap finish: timer = 196714 ms, PID=6879 - delta = 369ms
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-23-aneesh.kumar@linux.ibm.com
Now that all the lockless page table walk is careful w.r.t the PTE
address returned, we can now revert
commit: 13bd817bb8 ("powerpc/thp: Serialize pmd clear against a linux page table walk.")
We also drop the equivalent IPI from other pte updates routines. We still keep
IPI in hash pmdp collapse and that is to take care of parallel hash page table
insert. The radix pmdp collapse flush can possibly be removed once I am sure
generic code doesn't have the any expectations around parallel gup walk.
This speeds up Qemu guest RAM del/unplug time as below
128 core, 496GB guest:
Without patch:
munmap start: timer = 13162 ms, PID=7684
munmap finish: timer = 95312 ms, PID=7684 - delta = 82150 ms
With patch:
munmap start: timer = 196449 ms, PID=6681
munmap finish: timer = 196488 ms, PID=6681 - delta = 39ms
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-21-aneesh.kumar@linux.ibm.com
This adds _PAGE_PTE check and makes sure we validate the pte value returned via
find_kvm_host_pte.
NOTE: this also considers _PAGE_INVALID to the software valid bit.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-20-aneesh.kumar@linux.ibm.com
Current code just hold rmap lock to ensure parallel page table update is
prevented. That is not sufficient. The kernel should also check whether
a mmu_notifer callback was running in parallel.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-16-aneesh.kumar@linux.ibm.com
Since kvmppc_do_h_enter can get called in realmode use low level
arch_spin_lock which is safe to be called in realmode.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-15-aneesh.kumar@linux.ibm.com
The locking rules for walking nested shadow linux page table is different from process
scoped table. Hence add a helper for nested page table walk and also
add check whether we are holding the right locks.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-11-aneesh.kumar@linux.ibm.com
The locking rules for walking partition scoped table is different from process
scoped table. Hence add a helper for secondary linux page table walk and also
add check whether we are holding the right locks.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-10-aneesh.kumar@linux.ibm.com
These functions can get called in realmode. Hence use low level
arch_spin_lock which is safe to be called in realmode.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-9-aneesh.kumar@linux.ibm.com
read_user_stack_slow is called with interrupts soft disabled and it copies contents
from the page which we find mapped to a specific address. To convert
userspace address to pfn, the kernel now uses lockless page table walk.
The kernel needs to make sure the pfn value read remains stable and is not released
and reused for another process while the contents are read from the page. This
can only be achieved by holding a page reference.
One of the first approaches I tried was to check the pte value after the kernel
copies the contents from the page. But as shown below we can still get it wrong
CPU0 CPU1
pte = READ_ONCE(*ptep);
pte_clear(pte);
put_page(page);
page = alloc_page();
memcpy(page_address(page), "secret password", nr);
memcpy(buf, kaddr + offset, nb);
put_page(page);
handle_mm_fault()
page = alloc_page();
set_pte(pte, page);
if (pte_val(pte) != pte_val(*ptep))
Hence switch to __get_user_pages_fast.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-8-aneesh.kumar@linux.ibm.com
A lockless page table walk should be safe against parallel THP collapse, THP
split and madvise(MADV_DONTNEED)/parallel fault. This patch makes sure kernel
won't reload the pteval when checking for different conditions. The patch also added
a check for pte_present to make sure the kernel is indeed operating
on a PTE and not a pointer to level 0 table page.
The pfn value we find here can be different from the actual pfn on which
machine check happened. This can happen if we raced with a parallel update
of the page table. In such a scenario we end up isolating a wrong pfn. But that
doesn't have any other side effect.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-7-aneesh.kumar@linux.ibm.com
Don't fetch the pte value using lockless page table walk. Instead use the value from the
caller. hash_preload is called with ptl lock held. So it is safe to use the
pte_t address directly.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-6-aneesh.kumar@linux.ibm.com
This is only used with init_mm currently. Walking init_mm is much simpler
because we don't need to handle concurrent page table like other mm_context
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-5-aneesh.kumar@linux.ibm.com
This makes the pte_present check stricter by checking for additional _PAGE_PTE
bit. A level 1 pte pointer (THP pte) can be switched to a pointer to level 0 pte
page table page by following two operations.
1) THP split.
2) madvise(MADV_DONTNEED) in parallel to page fault.
A lockless page table walk need to make sure we can handle such changes
gracefully.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-4-aneesh.kumar@linux.ibm.com
If multiple threads in userspace keep changing the protection keys
mapping a range, there can be a scenario where kernel takes a key fault
but the pkey value found in the siginfo struct is a permissive one.
This can confuse the userspace as shown in the below test case.
/* use this to control the number of test iterations */
static void pkeyreg_set(int pkey, unsigned long rights)
{
unsigned long reg, shift;
shift = (NR_PKEYS - pkey - 1) * PKEY_BITS_PER_PKEY;
asm volatile("mfspr %0, 0xd" : "=r"(reg));
reg &= ~(((unsigned long) PKEY_BITS_MASK) << shift);
reg |= (rights & PKEY_BITS_MASK) << shift;
asm volatile("mtspr 0xd, %0" : : "r"(reg));
}
static unsigned long pkeyreg_get(void)
{
unsigned long reg;
asm volatile("mfspr %0, 0xd" : "=r"(reg));
return reg;
}
static int sys_pkey_mprotect(void *addr, size_t len, int prot, int pkey)
{
return syscall(SYS_pkey_mprotect, addr, len, prot, pkey);
}
static int sys_pkey_alloc(unsigned long flags, unsigned long access_rights)
{
return syscall(SYS_pkey_alloc, flags, access_rights);
}
static int sys_pkey_free(int pkey)
{
return syscall(SYS_pkey_free, pkey);
}
static int faulting_pkey;
static int permissive_pkey;
static pthread_barrier_t pkey_set_barrier;
static pthread_barrier_t mprotect_barrier;
static void pkey_handle_fault(int signum, siginfo_t *sinfo, void *ctx)
{
unsigned long pkeyreg;
/* FIXME: printf is not signal-safe but for the current purpose,
it gets the job done. */
printf("pkey: exp = %d, got = %d\n", faulting_pkey, sinfo->si_pkey);
fflush(stdout);
assert(sinfo->si_code == SEGV_PKUERR);
assert(sinfo->si_pkey == faulting_pkey);
/* clear pkey permissions to let the faulting instruction continue */
pkeyreg_set(faulting_pkey, 0x0);
}
static void *do_mprotect_fault(void *p)
{
unsigned long rights, pkeyreg, pgsize;
unsigned int i;
void *region;
int pkey;
srand(time(NULL));
pgsize = sysconf(_SC_PAGESIZE);
rights = PKEY_DISABLE_WRITE;
region = p;
/* allocate key, no permissions */
assert((pkey = sys_pkey_alloc(0, PKEY_DISABLE_ACCESS)) > 0);
pkeyreg_set(4, 0x0);
/* cache the pkey here as the faulting pkey for future reference
in the signal handler */
faulting_pkey = pkey;
printf("%s: faulting pkey = %d\n", __func__, faulting_pkey);
/* try to allocate, mprotect and free pkeys repeatedly */
for (i = 0; i < NUM_ITERATIONS; i++) {
/* sync up with the other thread here */
pthread_barrier_wait(&pkey_set_barrier);
/* make sure that the pkey used by the non-faulting thread
is made permissive for this thread's context too so that
no faults are triggered because it still might have been
set to a restrictive value */
// pkeyreg_set(permissive_pkey, 0x0);
/* sync up with the other thread here */
pthread_barrier_wait(&mprotect_barrier);
/* perform mprotect */
assert(!sys_pkey_mprotect(region, pgsize, PROT_READ | PROT_WRITE, pkey));
/* choose a random byte from the protected region and
attempt to write to it, this will generate a fault */
*((char *) region + (rand() % pgsize)) = rand();
/* restore pkey permissions as the signal handler may have
cleared the bit out for the sake of continuing */
pkeyreg_set(pkey, PKEY_DISABLE_WRITE);
}
/* free pkey */
sys_pkey_free(pkey);
return NULL;
}
static void *do_mprotect_nofault(void *p)
{
unsigned long pgsize;
unsigned int i, j;
void *region;
int pkey;
pgsize = sysconf(_SC_PAGESIZE);
region = p;
/* try to allocate, mprotect and free pkeys repeatedly */
for (i = 0; i < NUM_ITERATIONS; i++) {
/* allocate pkey, all permissions */
assert((pkey = sys_pkey_alloc(0, 0)) > 0);
permissive_pkey = pkey;
/* sync up with the other thread here */
pthread_barrier_wait(&pkey_set_barrier);
pthread_barrier_wait(&mprotect_barrier);
/* perform mprotect on the common page, no faults will
be triggered as this is most permissive */
assert(!sys_pkey_mprotect(region, pgsize, PROT_READ | PROT_WRITE, pkey));
/* free pkey */
assert(!sys_pkey_free(pkey));
}
return NULL;
}
int main(int argc, char **argv)
{
pthread_t fault_thread, nofault_thread;
unsigned long pgsize;
struct sigaction act;
pthread_attr_t attr;
cpu_set_t fault_cpuset, nofault_cpuset;
unsigned int i;
void *region;
/* allocate memory region to protect */
pgsize = sysconf(_SC_PAGESIZE);
assert(region = memalign(pgsize, pgsize));
CPU_ZERO(&fault_cpuset);
CPU_SET(0, &fault_cpuset);
CPU_ZERO(&nofault_cpuset);
CPU_SET(8, &nofault_cpuset);
assert(!pthread_attr_init(&attr));
/* setup sigsegv signal handler */
act.sa_handler = 0;
act.sa_sigaction = pkey_handle_fault;
assert(!sigprocmask(SIG_SETMASK, 0, &act.sa_mask));
act.sa_flags = SA_SIGINFO;
act.sa_restorer = 0;
assert(!sigaction(SIGSEGV, &act, NULL));
/* setup barrier for the two threads */
pthread_barrier_init(&pkey_set_barrier, NULL, 2);
pthread_barrier_init(&mprotect_barrier, NULL, 2);
/* setup and start threads */
assert(!pthread_create(&fault_thread, &attr, &do_mprotect_fault, region));
assert(!pthread_setaffinity_np(fault_thread, sizeof(cpu_set_t), &fault_cpuset));
assert(!pthread_create(&nofault_thread, &attr, &do_mprotect_nofault, region));
assert(!pthread_setaffinity_np(nofault_thread, sizeof(cpu_set_t), &nofault_cpuset));
/* cleanup */
assert(!pthread_attr_destroy(&attr));
assert(!pthread_join(fault_thread, NULL));
assert(!pthread_join(nofault_thread, NULL));
assert(!pthread_barrier_destroy(&pkey_set_barrier));
assert(!pthread_barrier_destroy(&mprotect_barrier));
free(region);
puts("PASS");
return EXIT_SUCCESS;
}
The above test can result the below failure without this patch.
pkey: exp = 3, got = 3
pkey: exp = 3, got = 4
a.out: pkey-siginfo-race.c💯 pkey_handle_fault: Assertion `sinfo->si_pkey == faulting_pkey' failed.
Aborted
Check for vma access before considering this a key fault. If vma pkey allow
access retry the acess again.
Test case is written by Sandipan Das <sandipan@linux.ibm.com> hence added SOB
from him.
Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-3-aneesh.kumar@linux.ibm.com
Fetch pkey from vma instead of linux page table. Also document the fact that in
some cases the pkey returned in siginfo won't be the same as the one we took
keyfault on. Even with linux page table walk, we can end up in a similar scenario.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200505071729.54912-2-aneesh.kumar@linux.ibm.com
- Fix a regression introduced in the last merge window, which results
in guests in HPT mode dying randomly.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQEcBAABCAAGBQJeni/pAAoJEJ2a6ncsY3GfTRoIANAQjIZi96AfJcfnrYQ4yUF7
scxawTiJ9VavvsEJLJ7vsozrJ4xxmvmA0fFWC84uw9+BwPqoLFFvZTjazbGEDVvF
FGwNBR/k7nfFVMIHS3K9iy9KjvYL3xkL26AgFTDJFq8hmOO9pH0txuk4r7SXb+NX
bGG0mScAD/Dg/HwAHAS6EP3jT35QtGTK62p8foqVTziTNcmBn9Ywtg0lEzAcq2iY
Y1BUD4Ov3cggshMI9SqHE8Yyq0XA2Wi6ggcyz/gVzvcbdFQmtg57Tri8nN8661LX
XKh+VTpYSIxNs5GgjwlNesJzJ9h6CSynJF556qrjQ0XsXcNqvn8fcZdNQ+hnRYw=
=Y19W
-----END PGP SIGNATURE-----
Merge tag 'kvm-ppc-fixes-5.7-1' into topic/ppc-kvm
This brings in a fix from the kvm-ppc tree that was merged to mainline
after rc2, and so isn't in the base of our topic branch. We'd like it
in the topic branch because it interacts with patches we plan to carry
in this branch.
Commit 0962e8004e ("powerpc/prom: Scan reserved-ranges node for
memory reservations") enabled support to parse reserved-ranges DT
node and reserve kernel memory falling in these ranges for F/W
purposes. Memory reserved for FADump should not overlap with these
ranges as it could corrupt memory meant for F/W or crash'ed kernel
memory to be exported as vmcore.
But since commit 579ca1a276 ("powerpc/fadump: make use of memblock's
bottom up allocation mode"), memblock_find_in_range() is being used to
find the appropriate area to reserve memory for FADump, which can't
account for reserved-ranges as these ranges are reserved only after
FADump memory reservation.
With reserved-ranges now being populated during early boot, look out
for these memory ranges while reserving memory for FADump. Without
this change, MPIPL on PowerNV systems aborts with hostboot failure,
when memory reserved for FADump is less than 4096MB.
Fixes: 579ca1a276 ("powerpc/fadump: make use of memblock's bottom up allocation mode")
Cc: stable@vger.kernel.org
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Reviewed-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/158737297693.26700.16193820746269425424.stgit@hbathini.in.ibm.com
At times, memory ranges have to be looked up during early boot, when
kernel couldn't be initialized for dynamic memory allocation. In fact,
reserved-ranges look up is needed during FADump memory reservation.
Without accounting for reserved-ranges in reserving memory for FADump,
MPIPL boot fails with memory corruption issues. So, extend memory
ranges handling to support static allocation and populate reserved
memory ranges during early boot.
Fixes: dda9dbfeeb ("powerpc/fadump: consider reserved ranges while releasing memory")
Cc: stable@vger.kernel.org
Signed-off-by: Hari Bathini <hbathini@linux.ibm.com>
Reviewed-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/158737294432.26700.4830263187856221314.stgit@hbathini.in.ibm.com
Move the static keyword to the front of declaration of 'vuart_bus_priv',
and resolve the following compiler warning that can be seen when
building with warnings enabled (W=1):
drivers/ps3/ps3-vuart.c:867:1: warning: ‘static’ is not at beginning of declaration [-Wold-style-declaration]
} static vuart_bus_priv;
^
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1588154448-56759-1-git-send-email-wangxiongfeng2@huawei.com
Currently, it is possible to have CONFIG_FUNCTION_TRACER disabled, but
CONFIG_MPROFILE_KERNEL enabled. Though all existing users of
MPROFILE_KERNEL are doing the right thing, it is weird to have
MPROFILE_KERNEL enabled when the function tracer isn't. Fix this by
making MPROFILE_KERNEL depend on FUNCTION_TRACER.
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200422092612.514301-1-naveen.n.rao@linux.vnet.ibm.com
Add documentation for the following sysfs interfaces:
/sys/devices/system/cpu/cpuX/purr
/sys/devices/system/cpu/cpuX/spurr
/sys/devices/system/cpu/cpuX/idle_purr
/sys/devices/system/cpu/cpuX/idle_spurr
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1586249263-14048-6-git-send-email-ego@linux.vnet.ibm.com
On Pseries LPARs, to calculate utilization, we need to know the
[S]PURR ticks when the CPUs were busy or idle.
The total PURR and SPURR ticks are already exposed via the per-cpu
sysfs files "purr" and "spurr". This patch adds support for exposing
the idle PURR and SPURR ticks via new per-cpu sysfs files named
"idle_purr" and "idle_spurr".
This patch also adds helper functions to accurately read the values of
idle_purr and idle_spurr especially from an interrupt context between
when the interrupt has occurred between the pseries_idle_prolog() and
pseries_idle_epilog(). This will ensure that the idle purr/spurr
values corresponding to the latest idle period is accounted for before
these values are read.
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1586249263-14048-5-git-send-email-ego@linux.vnet.ibm.com
On Pseries LPARs, to calculate utilization, we need to know the
[S]PURR ticks when the CPUs were busy or idle.
Via pseries_idle_prolog(), pseries_idle_epilog(), we track the idle
PURR ticks in the VPA variable "wait_state_cycles". This patch extends
the support to account for the idle SPURR ticks.
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1586249263-14048-4-git-send-email-ego@linux.vnet.ibm.com
Currently when CPU goes idle, we take a snapshot of PURR via
pseries_idle_prolog() which is used at the CPU idle exit to compute
the idle PURR cycles via the function pseries_idle_epilog(). Thus,
the value of idle PURR cycle thus read before pseries_idle_prolog() and
after pseries_idle_epilog() is always correct.
However, if we were to read the idle PURR cycles from an interrupt
context between pseries_idle_prolog() and pseries_idle_epilog() (this
will be done in a future patch), then, the value of the idle PURR thus
read will not include the cycles spent in the most recent idle period.
Thus, in that interrupt context, we will need access to the snapshot
of the PURR before going idle, in order to compute the idle PURR
cycles for the latest idle duration.
In this patch, we save the snapshot of PURR in pseries_idle_prolog()
in a per-cpu variable, instead of on the stack, so that it can be
accessed from an interrupt context.
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1586249263-14048-3-git-send-email-ego@linux.vnet.ibm.com
Currently prior to entering an idle state on a Linux Guest, the
pseries cpuidle driver implement an idle_loop_prolog() and
idle_loop_epilog() functions which ensure that idle_purr is correctly
computed, and the hypervisor is informed that the CPU cycles have been
donated.
These prolog and epilog functions are also required in the default
idle call, i.e pseries_lpar_idle(). Hence move these accessor
functions to a common header file and call them from
pseries_lpar_idle(). Since the existing header files such as
asm/processor.h have enough clutter, create a new header file
asm/idle.h. Finally rename idle_loop_prolog() and idle_loop_epilog()
to pseries_idle_prolog() and pseries_idle_epilog() as they are only
relavent for on pseries guests.
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1586249263-14048-2-git-send-email-ego@linux.vnet.ibm.com
allyesconfig fails with:
./usr/include/asm/vas-api.h:15:2: error: unknown type name '__u32'
15 | __u32 version;
| ^~~~~
./usr/include/asm/vas-api.h:16:2: error: unknown type name '__s16'
16 | __s16 vas_id; /* specific instance of vas or -1 for default */
| ^~~~~
./usr/include/asm/vas-api.h:17:2: error: unknown type name '__u16'
17 | __u16 reserved1;
| ^~~~~
./usr/include/asm/vas-api.h:18:2: error: unknown type name '__u64'
18 | __u64 flags; /* Future use */
| ^~~~~
./usr/include/asm/vas-api.h:19:2: error: unknown type name '__u64'
19 | __u64 reserved2[6];
| ^~~~~
uapi headers should be self contained, so add an include of
linux/types.h.
Fixes: 45f25a79fe ("powerpc/vas: Define VAS_TX_WIN_OPEN ioctl API")
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Haren Myneni <haren@linux.ibm.com>
[mpe: Flesh out change log from linux-next error report]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200422154129.11f988fd@canb.auug.org.au
Include a README file with the instructions to use the
testcases at selftests/powerpc/nx-gzip.
Signed-off-by: Bulent Abali <abali@us.ibm.com>
Signed-off-by: Raphael Moreira Zinsly <rzinsly@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200420205538.25181-6-rzinsly@linux.ibm.com
Include a decompression testcase for the powerpc NX-GZIP
engine.
Signed-off-by: Bulent Abali <abali@us.ibm.com>
Signed-off-by: Raphael Moreira Zinsly <rzinsly@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200420205538.25181-5-rzinsly@linux.ibm.com
Add files to be able to compress and decompress files using the
powerpc NX-GZIP engine.
Signed-off-by: Bulent Abali <abali@us.ibm.com>
Signed-off-by: Raphael Moreira Zinsly <rzinsly@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200420205538.25181-3-rzinsly@linux.ibm.com
As described by Haren:
Power9 processor supports Virtual Accelerator Switchboard (VAS) which
allows kernel and userspace to send compression requests to Nest
Accelerator (NX) directly. The NX unit comprises of 2 842 compression
engines and 1 GZIP engine. Linux kernel already has 842 compression
support on kernel. This patch series adds GZIP compression support
from user space. The GZIP Compression engine implements the ZLIB and
GZIP compression algorithms. No plans of adding NX-GZIP compression
support in kernel right now.
Applications can send requests to NX directly with COPY/PASTE
instructions. But kernel has to establish channel / window on NX-GZIP
device for the userspace. So userspace access to the GZIP engine is
provided through /dev/crypto/nx-gzip device with several operations.
An application must open the this device to obtain a file
descriptor (fd). Using the fd, application should issue the
VAS_TX_WIN_OPEN ioctl to establish a connection to the engine. Once
window is opened, should use mmap() system call to map the hardware
address of engine's request queue into the application's virtual
address space. Then user space forms the request as co-processor
Request Block (CRB) and paste this CRB on the mapped HW address using
COPY/PASTE instructions. Application can poll on status flags (part of
CRB) with timeout for request completion.
For VAS_TX_WIN_OPEN ioctl, if user space passes vas_id = -1 (struct
vas_tx_win_open_attr), kernel determines the VAS instance on the
corresponding chip based on the CPU on which the process is executing.
Otherwise, the specified VAS instance is used if application passes
the proper VAS instance (vas_id listed in
/proc/device-tree/vas@*/ibm,vas_id).
Process can open multiple windows with different FDs or can send
several requests to NX on the same window at the same time.
As described by Haren:
On Power9, Virtual Accelerator Switchboard (VAS) allows user space or
kernel to communicate with Nest Accelerator (NX) directly using
COPY/PASTE instructions. NX provides various functionalities such as
compression, encryption and etc. But only compression (842 and GZIP
formats) is supported in Linux kernel on power9.
842 compression driver (drivers/crypto/nx/nx-842-powernv.c) is already
included in Linux. Only GZIP support will be available from user
space.
Applications can issue GZIP compression / decompression requests to NX
with COPY/PASTE instructions. When NX is processing these requests,
can hit fault on the request buffer (not in memory). It issues an
interrupt and pastes fault CRB in fault FIFO. Expects kernel to handle
this fault and return credits for both send and fault windows after
processing.
This patch series adds IRQ and fault window setup, and NX fault
handling:
- Alloc IRQ and trigger port address, and configure IRQ per VAS
instance.
- Set port# for each window to generate an interrupt when noticed
fault.
- Set fault window and FIFO on which NX paste fault CRB.
- Setup IRQ thread fault handler per VAS instance.
- When receiving an interrupt, Read CRBs from fault FIFO and update
coprocessor_status_block (CSB) in the corresponding CRB with
translation failure (CSB_CC_TRANSLATION). After issuing NX
requests, process polls on CSB address. When it sees translation
error, can touch the request buffer to bring the page in to memory
and reissue NX request.
- If copy_to_user fails on user space CSB address, OS sends SEGV
signal.
Since cd758a9b57 "KVM: PPC: Book3S HV: Use __gfn_to_pfn_memslot in HPT
page fault handler", it's been possible in fairly rare circumstances to
load a non-present PTE in kvmppc_book3s_hv_page_fault() when running a
guest on a POWER8 host.
Because that case wasn't checked for, we could misinterpret the non-present
PTE as being a cache-inhibited PTE. That could mismatch with the
corresponding hash PTE, which would cause the function to fail with -EFAULT
a little further down. That would propagate up to the KVM_RUN ioctl()
generally causing the KVM userspace (usually qemu) to fall over.
This addresses the problem by catching that case and returning to the guest
instead.
For completeness, this fixes the radix page fault handler in the same
way. For radix this didn't cause any obvious misbehaviour, because we
ended up putting the non-present PTE into the guest's partition-scoped
page tables, leading immediately to another hypervisor data/instruction
storage interrupt, which would go through the page fault path again
and fix things up.
Fixes: cd758a9b57 "KVM: PPC: Book3S HV: Use __gfn_to_pfn_memslot in HPT page fault handler"
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1820402
Reported-by: David Gibson <david@gibson.dropbear.id.au>
Tested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Power9 introduced Virtual Accelerator Switchboard (VAS) which allows
userspace to communicate with Nest Accelerator (NX) directly. But
kernel has to establish channel to NX for userspace. This document
describes user space API that application can use to establish
communication channel.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114715.2275.1135.camel@hbabu-laptop
When window is opened, pid reference is taken for user space
windows. Not needed for kernel windows. So remove 'pid' in
vas_tx_win_attr struct.
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114674.2275.1132.camel@hbabu-laptop
Changes to probe GZIP device-tree nodes, open RX windows and setup
GZIP compression type. No plans to provide GZIP usage in kernel right
now, but this patch enables GZIP for user space usage.
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114624.2275.1129.camel@hbabu-laptop
Make setup and enable code generic to support new GZIP compression type.
Changed nx842 reference to nx and moved some code to new functions.
Functionality is not changed except sparse warning fix - setting NULL
instead of 0 for per_cpu send window in nx_delete_coprocs().
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114569.2275.1126.camel@hbabu-laptop
Rename nx-842-powernv.c to nx-common-powernv.c to add code for setup
and enable new GZIP compression type. The actual functionality is not
changed in this patch.
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114200.2275.1113.camel@hbabu-laptop
coproc entry is initialized during NX probe on power9, but not on P8.
nx842_delete_coprocs() is used for both and frees receive window if it
is allocated. Getting crash for rmmod on P8 since coproc->vas.rxwin
is not initialized.
This patch replaces kmalloc with kzalloc in nx842_powernv_probe()
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114163.2275.1110.camel@hbabu-laptop
On power9, userspace can send GZIP compression requests directly to NX
once kernel establishes NX channel / window with VAS. This patch provides
user space API which allows user space to establish channel using open
VAS_TX_WIN_OPEN ioctl, mmap and close operations.
Each window corresponds to file descriptor and application can open
multiple windows. After the window is opened, VAS_TX_WIN_OPEN icoctl to
open a window on specific VAS instance, mmap() system call to map
the hardware address of engine's request queue into the application's
virtual address space.
Then the application can then submit one or more requests to the the
engine by using the copy/paste instructions and pasting the CRBs to
the virtual address (aka paste_address) returned by mmap().
Only NX GZIP coprocessor type is supported right now and allow GZIP
engine access via /dev/crypto/nx-gzip device node.
Thanks to Michael Ellerman for his changes and suggestions to make the
ioctl generic to support any coprocessor type.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Haren Myneni <haren@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1587114121.2275.1109.camel@hbabu-laptop
Christophe JAILLET