Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'powerpc-5.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"One fix for a regression introduced by our 32-bit KASAN support, which
broke booting on machines with "bootx" early debugging enabled.
A fix for a bug which broke kexec on 32-bit, introduced by changes to
the 32-bit STRICT_KERNEL_RWX support in v5.1.
Finally two fixes going to stable for our THP split/collapse handling,
discovered by Nick. The first fixes random crashes and/or corruption
in guests under sufficient load.
Thanks to: Nicholas Piggin, Christophe Leroy, Aaro Koskinen, Mathieu
Malaterre"
* tag 'powerpc-5.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/32s: fix booting with CONFIG_PPC_EARLY_DEBUG_BOOTX
powerpc/64s: __find_linux_pte() synchronization vs pmdp_invalidate()
powerpc/64s: Fix THP PMD collapse serialisation
powerpc: Fix kexec failure on book3s/32
+59
-4
+30
arch/powerpc/include/asm/book3s/64/pgtable.h
+30
arch/powerpc/include/asm/book3s/64/pgtable.h
···
876
876
return false;
877
877
}
878
878
879
+
static inline int pmd_is_serializing(pmd_t pmd)
880
+
{
881
+
/*
882
+
* If the pmd is undergoing a split, the _PAGE_PRESENT bit is clear
883
+
* and _PAGE_INVALID is set (see pmd_present, pmdp_invalidate).
884
+
*
885
+
* This condition may also occur when flushing a pmd while flushing
886
+
* it (see ptep_modify_prot_start), so callers must ensure this
887
+
* case is fine as well.
888
+
*/
889
+
if ((pmd_raw(pmd) & cpu_to_be64(_PAGE_PRESENT | _PAGE_INVALID)) ==
890
+
cpu_to_be64(_PAGE_INVALID))
891
+
return true;
892
+
893
+
return false;
894
+
}
895
+
879
896
static inline int pmd_bad(pmd_t pmd)
880
897
{
881
898
if (radix_enabled())
···
1109
1092
#define pmd_access_permitted pmd_access_permitted
1110
1093
static inline bool pmd_access_permitted(pmd_t pmd, bool write)
1111
1094
{
1095
+
/*
1096
+
* pmdp_invalidate sets this combination (which is not caught by
1097
+
* !pte_present() check in pte_access_permitted), to prevent
1098
+
* lock-free lookups, as part of the serialize_against_pte_lookup()
1099
+
* synchronisation.
1100
+
*
1101
+
* This also catches the case where the PTE's hardware PRESENT bit is
1102
+
* cleared while TLB is flushed, which is suboptimal but should not
1103
+
* be frequent.
1104
+
*/
1105
+
if (pmd_is_serializing(pmd))
1106
+
return false;
1107
+
1112
1108
return pte_access_permitted(pmd_pte(pmd), write);
1113
1109
}
1114
1110
+4
arch/powerpc/include/asm/btext.h
+4
arch/powerpc/include/asm/btext.h
···
13
13
int depth, int pitch);
14
14
extern void btext_setup_display(int width, int height, int depth, int pitch,
15
15
unsigned long address);
16
+
#ifdef CONFIG_PPC32
16
17
extern void btext_prepare_BAT(void);
18
+
#else
19
+
static inline void btext_prepare_BAT(void) { }
20
+
#endif
17
21
extern void btext_map(void);
18
22
extern void btext_unmap(void);
19
23
+3
arch/powerpc/include/asm/kexec.h
+3
arch/powerpc/include/asm/kexec.h
+3
-1
arch/powerpc/kernel/machine_kexec_32.c
+3
-1
arch/powerpc/kernel/machine_kexec_32.c
···
30
30
*/
31
31
void default_machine_kexec(struct kimage *image)
32
32
{
33
-
extern const unsigned char relocate_new_kernel[];
34
33
extern const unsigned int relocate_new_kernel_size;
35
34
unsigned long page_list;
36
35
unsigned long reboot_code_buffer, reboot_code_buffer_phys;
···
56
57
flush_icache_range(reboot_code_buffer,
57
58
reboot_code_buffer + KEXEC_CONTROL_PAGE_SIZE);
58
59
printk(KERN_INFO "Bye!\n");
60
+
61
+
if (!IS_ENABLED(CONFIG_FSL_BOOKE) && !IS_ENABLED(CONFIG_44x))
62
+
relocate_new_kernel(page_list, reboot_code_buffer_phys, image->start);
59
63
60
64
/* now call it */
61
65
rnk = (relocate_new_kernel_t) reboot_code_buffer;
+1
arch/powerpc/kernel/prom_init.c
+1
arch/powerpc/kernel/prom_init.c
+1
-1
arch/powerpc/kernel/prom_init_check.sh
+1
-1
arch/powerpc/kernel/prom_init_check.sh
···
24
24
WHITELIST="add_reloc_offset __bss_start __bss_stop copy_and_flush
25
25
_end enter_prom $MEM_FUNCS reloc_offset __secondary_hold
26
26
__secondary_hold_acknowledge __secondary_hold_spinloop __start
27
-
logo_linux_clut224
27
+
logo_linux_clut224 btext_prepare_BAT
28
28
reloc_got2 kernstart_addr memstart_addr linux_banner _stext
29
29
__prom_init_toc_start __prom_init_toc_end btext_setup_display TOC."
30
30
+3
arch/powerpc/mm/book3s64/pgtable.c
+3
arch/powerpc/mm/book3s64/pgtable.c
···
112
112
/*
113
113
* This ensures that generic code that rely on IRQ disabling
114
114
* to prevent a parallel THP split work as expected.
115
+
*
116
+
* Marking the entry with _PAGE_INVALID && ~_PAGE_PRESENT requires
117
+
* a special case check in pmd_access_permitted.
115
118
*/
116
119
serialize_against_pte_lookup(vma->vm_mm);
117
120
return __pmd(old_pmd);
+14
-2
arch/powerpc/mm/pgtable.c
+14
-2
arch/powerpc/mm/pgtable.c
···
368
368
pdshift = PMD_SHIFT;
369
369
pmdp = pmd_offset(&pud, ea);
370
370
pmd = READ_ONCE(*pmdp);
371
+
371
372
/*
372
-
* A hugepage collapse is captured by pmd_none, because
373
-
* it mark the pmd none and do a hpte invalidate.
373
+
* A hugepage collapse is captured by this condition, see
374
+
* pmdp_collapse_flush.
374
375
*/
375
376
if (pmd_none(pmd))
376
377
return NULL;
378
+
379
+
#ifdef CONFIG_PPC_BOOK3S_64
380
+
/*
381
+
* A hugepage split is captured by this condition, see
382
+
* pmdp_invalidate.
383
+
*
384
+
* Huge page modification can be caught here too.
385
+
*/
386
+
if (pmd_is_serializing(pmd))
387
+
return NULL;
388
+
#endif
377
389
378
390
if (pmd_trans_huge(pmd) || pmd_devmap(pmd)) {
379
391
if (is_thp)