Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Le 29/10/2020 à 22:07, Andreas Schwab a écrit : On Okt 01 2020, Christophe Leroy wrote: At the time being, an early hash table is set up when CONFIG_KASAN is selected. There is nothing wrong with setting such an early hash table all the time, even if it is not used. This is a statically allocated 256 kB table which lies in the init data section. This makes the code simpler and may in the future allow to setup early IO mappings with fixmap instead of hard coding BATs. Put create_hpte() and flush_hash_pages() in the .ref.text section in order to avoid warning for the reference to early_hash[]. This reference is removed by MMU_init_hw_patch() before init memory is freed. This breaks booting on the iBook G4. Can you test patch https://patchwork.ozlabs.org/project/linuxppc-dev/patch/9e225a856a8b22e0e77587ee22ab7a2f5bca8753.1604740029.git.christophe.le...@csgroup.eu/ Thanks Christophe
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
>>> To be sure we are not in front of a long lasting bug, could you try >>> CONFIG_KASAN=y on v5.9 ? >> >> Indeed it started to fail somewhere between v5.6 and v5.7. >> >> v5.7 fails early with few messages on the console with reboot, v5.8 and >> later hang right at bootloader. >> >> I'm bisecting now. > > (side note: I tried FB_OF=y instead of DRM_RADEON + DRM_FBDEV_* to speed > up bisection and it turns out in that configuration KASAN never worked, > down to commit 305d600123046, hanging right after bootloader or even > with invalid access in the bootloader itself). My bisection ended up nowhere (at net-next merge with 2k commits), and given the above failure with unrelated configuration change, I conclude that KASAN=y was always broken on this box.
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
>> To be sure we are not in front of a long lasting bug, could you try >> CONFIG_KASAN=y on v5.9 ? > > Indeed it started to fail somewhere between v5.6 and v5.7. > > v5.7 fails early with few messages on the console with reboot, v5.8 and > later hang right at bootloader. > > I'm bisecting now. (side note: I tried FB_OF=y instead of DRM_RADEON + DRM_FBDEV_* to speed up bisection and it turns out in that configuration KASAN never worked, down to commit 305d600123046, hanging right after bootloader or even with invalid access in the bootloader itself).
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Christophe Leroy writes: > To be sure we are not in front of a long lasting bug, could you try > CONFIG_KASAN=y on v5.9 ? Indeed it started to fail somewhere between v5.6 and v5.7. v5.7 fails early with few messages on the console with reboot, v5.8 and later hang right at bootloader. I'm bisecting now.
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Le 03/11/2020 à 19:58, Serge Belyshev a écrit : Would you mind checking that with that patch reverted, you are able to boot a kernel built with CONFIG_KASAN ? I can reproduce the same problem on a powerbook G4, and no, CONFIG_KASAN=y kernel with that patch reverted also does not boot with the same symptom: white screen at the bootloader right after "Booting Linux via __start() @ 0x014 ..." Thanks for the test Serge. To be sure we are not in front of a long lasting bug, could you try CONFIG_KASAN=y on v5.9 ? Christophe
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
> Would you mind checking that with that patch reverted, you are able to > boot a kernel built with CONFIG_KASAN ? I can reproduce the same problem on a powerbook G4, and no, CONFIG_KASAN=y kernel with that patch reverted also does not boot with the same symptom: white screen at the bootloader right after "Booting Linux via __start() @ 0x014 ..."
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
On Nov 03 2020, Christophe Leroy wrote: > Would you mind checking that with that patch reverted, you are able to > boot a kernel built with CONFIG_KASAN ? That doesn't exist. Andreas. -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
On Nov 03 2020, Christophe Leroy wrote: > I tried again on QEMU with both pmac32_defconfig and your config, and it > boots. Isn't it quite naïve to think that qemu provides an adequate test environment for such lowlevel stuff? Andreas. -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Hi Andreas, Le 30/10/2020 à 14:11, Andreas Schwab a écrit : # # Automatically generated file; DO NOT EDIT. # Linux/powerpc 5.10.0-rc1 Kernel Configuration # I tried again on QEMU with both pmac32_defconfig and your config, and it boots. I really can't understand what the problem is, because that patch only activates at all time something that has been working well when CONFIG_KASAN is set. Would you mind checking that with that patch reverted, you are able to boot a kernel built with CONFIG_KASAN ? Thanks Christophe
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
# # Automatically generated file; DO NOT EDIT. # Linux/powerpc 5.10.0-rc1 Kernel Configuration # CONFIG_CC_VERSION_TEXT="gcc-4.9 (SUSE Linux) 4.9.3" CONFIG_CC_IS_GCC=y CONFIG_GCC_VERSION=40903 CONFIG_LD_VERSION=23501 CONFIG_CLANG_VERSION=0 CONFIG_CC_CAN_LINK=y CONFIG_CC_CAN_LINK_STATIC=y CONFIG_CC_HAS_ASM_GOTO=y CONFIG_IRQ_WORK=y CONFIG_BUILDTIME_TABLE_SORT=y CONFIG_THREAD_INFO_IN_TASK=y # # General setup # CONFIG_BROKEN_ON_SMP=y CONFIG_INIT_ENV_ARG_LIMIT=32 # CONFIG_COMPILE_TEST is not set CONFIG_LOCALVERSION="" # CONFIG_LOCALVERSION_AUTO is not set CONFIG_BUILD_SALT="" CONFIG_HAVE_KERNEL_GZIP=y CONFIG_HAVE_KERNEL_XZ=y CONFIG_KERNEL_GZIP=y # CONFIG_KERNEL_XZ is not set CONFIG_DEFAULT_INIT="" CONFIG_DEFAULT_HOSTNAME="(none)" CONFIG_SWAP=y CONFIG_SYSVIPC=y CONFIG_SYSVIPC_SYSCTL=y CONFIG_POSIX_MQUEUE=y CONFIG_POSIX_MQUEUE_SYSCTL=y CONFIG_WATCH_QUEUE=y CONFIG_CROSS_MEMORY_ATTACH=y CONFIG_USELIB=y CONFIG_AUDIT=y CONFIG_HAVE_ARCH_AUDITSYSCALL=y CONFIG_AUDITSYSCALL=y # # IRQ subsystem # CONFIG_GENERIC_IRQ_SHOW=y CONFIG_GENERIC_IRQ_SHOW_LEVEL=y CONFIG_IRQ_DOMAIN=y CONFIG_IRQ_FORCED_THREADING=y CONFIG_SPARSE_IRQ=y # CONFIG_GENERIC_IRQ_DEBUGFS is not set # end of IRQ subsystem CONFIG_GENERIC_TIME_VSYSCALL=y CONFIG_GENERIC_CLOCKEVENTS=y CONFIG_GENERIC_CMOS_UPDATE=y # # Timers subsystem # CONFIG_TICK_ONESHOT=y CONFIG_NO_HZ_COMMON=y # CONFIG_HZ_PERIODIC is not set CONFIG_NO_HZ_IDLE=y CONFIG_NO_HZ=y CONFIG_HIGH_RES_TIMERS=y # end of Timers subsystem # CONFIG_PREEMPT_NONE is not set CONFIG_PREEMPT_VOLUNTARY=y # CONFIG_PREEMPT is not set # # CPU/Task time and stats accounting # CONFIG_TICK_CPU_ACCOUNTING=y # CONFIG_VIRT_CPU_ACCOUNTING_NATIVE is not set # CONFIG_IRQ_TIME_ACCOUNTING is not set CONFIG_BSD_PROCESS_ACCT=y CONFIG_BSD_PROCESS_ACCT_V3=y CONFIG_TASKSTATS=y CONFIG_TASK_DELAY_ACCT=y CONFIG_TASK_XACCT=y CONFIG_TASK_IO_ACCOUNTING=y CONFIG_PSI=y # CONFIG_PSI_DEFAULT_DISABLED is not set # end of CPU/Task time and stats accounting # # RCU Subsystem # CONFIG_TINY_RCU=y # CONFIG_RCU_EXPERT is not set CONFIG_SRCU=y CONFIG_TINY_SRCU=y CONFIG_TASKS_RCU_GENERIC=y CONFIG_TASKS_TRACE_RCU=y # end of RCU Subsystem CONFIG_IKCONFIG=y CONFIG_IKCONFIG_PROC=y # CONFIG_IKHEADERS is not set CONFIG_LOG_BUF_SHIFT=18 CONFIG_PRINTK_SAFE_LOG_BUF_SHIFT=13 # # Scheduler features # # end of Scheduler features CONFIG_CGROUPS=y CONFIG_PAGE_COUNTER=y CONFIG_MEMCG=y CONFIG_MEMCG_SWAP=y CONFIG_MEMCG_KMEM=y CONFIG_BLK_CGROUP=y CONFIG_CGROUP_WRITEBACK=y CONFIG_CGROUP_SCHED=y CONFIG_FAIR_GROUP_SCHED=y CONFIG_CFS_BANDWIDTH=y CONFIG_RT_GROUP_SCHED=y CONFIG_CGROUP_PIDS=y # CONFIG_CGROUP_RDMA is not set CONFIG_CGROUP_FREEZER=y CONFIG_CGROUP_DEVICE=y CONFIG_CGROUP_CPUACCT=y # CONFIG_CGROUP_BPF is not set # CONFIG_CGROUP_DEBUG is not set CONFIG_SOCK_CGROUP_DATA=y CONFIG_NAMESPACES=y CONFIG_UTS_NS=y CONFIG_IPC_NS=y CONFIG_USER_NS=y CONFIG_PID_NS=y CONFIG_NET_NS=y # CONFIG_CHECKPOINT_RESTORE is not set CONFIG_SCHED_AUTOGROUP=y # CONFIG_SYSFS_DEPRECATED is not set CONFIG_RELAY=y CONFIG_BLK_DEV_INITRD=y CONFIG_INITRAMFS_SOURCE="" CONFIG_RD_GZIP=y CONFIG_RD_BZIP2=y CONFIG_RD_LZMA=y CONFIG_RD_XZ=y CONFIG_RD_LZO=y CONFIG_RD_LZ4=y CONFIG_RD_ZSTD=y CONFIG_BOOT_CONFIG=y # CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE is not set CONFIG_CC_OPTIMIZE_FOR_SIZE=y CONFIG_HAVE_LD_DEAD_CODE_DATA_ELIMINATION=y # CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is not set CONFIG_SYSCTL=y CONFIG_SYSCTL_EXCEPTION_TRACE=y CONFIG_BPF=y CONFIG_EXPERT=y CONFIG_MULTIUSER=y CONFIG_SGETMASK_SYSCALL=y CONFIG_SYSFS_SYSCALL=y CONFIG_FHANDLE=y CONFIG_POSIX_TIMERS=y CONFIG_PRINTK=y CONFIG_BUG=y CONFIG_ELF_CORE=y CONFIG_BASE_FULL=y CONFIG_FUTEX=y CONFIG_FUTEX_PI=y CONFIG_EPOLL=y CONFIG_SIGNALFD=y CONFIG_TIMERFD=y CONFIG_EVENTFD=y CONFIG_SHMEM=y CONFIG_AIO=y CONFIG_IO_URING=y CONFIG_ADVISE_SYSCALLS=y CONFIG_MEMBARRIER=y CONFIG_KALLSYMS=y CONFIG_KALLSYMS_ALL=y CONFIG_KALLSYMS_BASE_RELATIVE=y CONFIG_BPF_SYSCALL=y # CONFIG_BPF_PRELOAD is not set CONFIG_USERFAULTFD=y CONFIG_ARCH_HAS_MEMBARRIER_CALLBACKS=y CONFIG_ARCH_HAS_MEMBARRIER_SYNC_CORE=y CONFIG_RSEQ=y # CONFIG_DEBUG_RSEQ is not set # CONFIG_EMBEDDED is not set CONFIG_HAVE_PERF_EVENTS=y # CONFIG_PC104 is not set # # Kernel Performance Events And Counters # # CONFIG_PERF_EVENTS is not set # end of Kernel Performance Events And Counters CONFIG_VM_EVENT_COUNTERS=y CONFIG_SLUB_DEBUG=y CONFIG_SLUB_MEMCG_SYSFS_ON=y # CONFIG_COMPAT_BRK is not set # CONFIG_SLAB is not set CONFIG_SLUB=y # CONFIG_SLOB is not set CONFIG_SLAB_MERGE_DEFAULT=y # CONFIG_SLAB_FREELIST_RANDOM is not set # CONFIG_SLAB_FREELIST_HARDENED is not set # CONFIG_SHUFFLE_PAGE_ALLOCATOR is not set CONFIG_SYSTEM_DATA_VERIFICATION=y # CONFIG_PROFILING is not set # end of General setup CONFIG_PPC32=y # CONFIG_PPC64 is not set CONFIG_PPC_BOOK3S_32=y # # Processor support # CONFIG_PPC_BOOK3S_6xx=y # CONFIG_PPC_85xx is not set # CONFIG_PPC_8xx is not set # CONFIG_40x is not set # CONFIG_44x is not set # CONFIG_E200 is not set # CONFIG_GENERIC_CPU is not set # CONFIG_E300C2_CPU is not set #
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Andreas Schwab a écrit : On Okt 30 2020, Michael Ellerman wrote: Andreas Schwab writes: On Okt 01 2020, Christophe Leroy wrote: At the time being, an early hash table is set up when CONFIG_KASAN is selected. There is nothing wrong with setting such an early hash table all the time, even if it is not used. This is a statically allocated 256 kB table which lies in the init data section. This makes the code simpler and may in the future allow to setup early IO mappings with fixmap instead of hard coding BATs. Put create_hpte() and flush_hash_pages() in the .ref.text section in order to avoid warning for the reference to early_hash[]. This reference is removed by MMU_init_hw_patch() before init memory is freed. This breaks booting on the iBook G4. Do you get an oops or anything? Nope, nothing at all. Under qemu it was working. Can you share your .config, or are you using one of the defconfig ? Christophe
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
On Okt 30 2020, Michael Ellerman wrote: > Andreas Schwab writes: >> On Okt 01 2020, Christophe Leroy wrote: >> >>> At the time being, an early hash table is set up when >>> CONFIG_KASAN is selected. >>> >>> There is nothing wrong with setting such an early hash table >>> all the time, even if it is not used. This is a statically >>> allocated 256 kB table which lies in the init data section. >>> >>> This makes the code simpler and may in the future allow to >>> setup early IO mappings with fixmap instead of hard coding BATs. >>> >>> Put create_hpte() and flush_hash_pages() in the .ref.text section >>> in order to avoid warning for the reference to early_hash[]. This >>> reference is removed by MMU_init_hw_patch() before init memory is >>> freed. >> >> This breaks booting on the iBook G4. > > Do you get an oops or anything? Nope, nothing at all. Andreas. -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
Andreas Schwab writes: > On Okt 01 2020, Christophe Leroy wrote: > >> At the time being, an early hash table is set up when >> CONFIG_KASAN is selected. >> >> There is nothing wrong with setting such an early hash table >> all the time, even if it is not used. This is a statically >> allocated 256 kB table which lies in the init data section. >> >> This makes the code simpler and may in the future allow to >> setup early IO mappings with fixmap instead of hard coding BATs. >> >> Put create_hpte() and flush_hash_pages() in the .ref.text section >> in order to avoid warning for the reference to early_hash[]. This >> reference is removed by MMU_init_hw_patch() before init memory is >> freed. > > This breaks booting on the iBook G4. Do you get an oops or anything? cheers
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
On Okt 01 2020, Christophe Leroy wrote: > At the time being, an early hash table is set up when > CONFIG_KASAN is selected. > > There is nothing wrong with setting such an early hash table > all the time, even if it is not used. This is a statically > allocated 256 kB table which lies in the init data section. > > This makes the code simpler and may in the future allow to > setup early IO mappings with fixmap instead of hard coding BATs. > > Put create_hpte() and flush_hash_pages() in the .ref.text section > in order to avoid warning for the reference to early_hash[]. This > reference is removed by MMU_init_hw_patch() before init memory is > freed. This breaks booting on the iBook G4. Andreas. -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."
Re: [PATCH] powerpc/32s: Setup the early hash table at all time.
On Thu, 1 Oct 2020 15:35:38 + (UTC), Christophe Leroy wrote: > At the time being, an early hash table is set up when > CONFIG_KASAN is selected. > > There is nothing wrong with setting such an early hash table > all the time, even if it is not used. This is a statically > allocated 256 kB table which lies in the init data section. > > [...] Applied to powerpc/next. [1/1] powerpc/32s: Setup the early hash table at all time. https://git.kernel.org/powerpc/c/69a1593abdbcf03a76367320d929a8ae7a5e3d71 cheers
[PATCH] powerpc/32s: Setup the early hash table at all time.
At the time being, an early hash table is set up when
CONFIG_KASAN is selected.
There is nothing wrong with setting such an early hash table
all the time, even if it is not used. This is a statically
allocated 256 kB table which lies in the init data section.
This makes the code simpler and may in the future allow to
setup early IO mappings with fixmap instead of hard coding BATs.
Put create_hpte() and flush_hash_pages() in the .ref.text section
in order to avoid warning for the reference to early_hash[]. This
reference is removed by MMU_init_hw_patch() before init memory is
freed.
Signed-off-by: Christophe Leroy
---
arch/powerpc/kernel/head_32.S | 13 +
arch/powerpc/mm/book3s32/hash_low.S | 9 +++--
arch/powerpc/mm/book3s32/mmu.c| 14 +-
arch/powerpc/mm/kasan/kasan_init_32.c | 19 ---
4 files changed, 17 insertions(+), 38 deletions(-)
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 2bd0aa3a4cc7..b5458113e0b0 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -166,9 +166,9 @@ __after_mmu_off:
bl initial_bats
bl load_segment_registers
-#ifdef CONFIG_KASAN
+BEGIN_MMU_FTR_SECTION
bl early_hash_table
-#endif
+END_MMU_FTR_SECTION_IFSET(MMU_FTR_HPTE_TABLE)
#if defined(CONFIG_BOOTX_TEXT)
bl setup_disp_bat
#endif
@@ -953,7 +953,6 @@ _ENTRY(__restore_cpu_setup)
* Load stuff into the MMU. Intended to be called with
* IR=0 and DR=0.
*/
-#ifdef CONFIG_KASAN
early_hash_table:
sync/* Force all PTE updates to finish */
isync
@@ -964,8 +963,10 @@ early_hash_table:
lis r6, early_hash - PAGE_OFFSET@h
ori r6, r6, 3 /* 256kB table */
mtspr SPRN_SDR1, r6
+ lis r6, early_hash@h
+ lis r3, Hash@ha
+ stw r6, Hash@l(r3)
blr
-#endif
load_up_mmu:
sync/* Force all PTE updates to finish */
@@ -1055,11 +1056,7 @@ start_here:
bl machine_init
bl __save_cpu_setup
bl MMU_init
-#ifdef CONFIG_KASAN
-BEGIN_MMU_FTR_SECTION
bl MMU_init_hw_patch
-END_MMU_FTR_SECTION_IFSET(MMU_FTR_HPTE_TABLE)
-#endif
/*
* Go back to running unmapped so we can load up new values
diff --git a/arch/powerpc/mm/book3s32/hash_low.S
b/arch/powerpc/mm/book3s32/hash_low.S
index 1690d369688b..8fc594ff7286 100644
--- a/arch/powerpc/mm/book3s32/hash_low.S
+++ b/arch/powerpc/mm/book3s32/hash_low.S
@@ -15,6 +15,7 @@
*/
#include
+#include
#include
#include
#include
@@ -287,9 +288,9 @@ _ASM_NOKPROBE_SYMBOL(add_hash_page)
*
* For speed, 4 of the instructions get patched once the size and
* physical address of the hash table are known. These definitions
- * of Hash_base and Hash_bits below are just an example.
+ * of Hash_base and Hash_bits below are for the early hash table.
*/
-Hash_base = 0xc018
+Hash_base = early_hash
Hash_bits = 12 /* e.g. 256kB hash table */
Hash_msk = (((1 << Hash_bits) - 1) * 64)
@@ -310,6 +311,7 @@ Hash_msk = (((1 << Hash_bits) - 1) * 64)
#define HASH_LEFT 31-(LG_PTEG_SIZE+Hash_bits-1)
#define HASH_RIGHT 31-LG_PTEG_SIZE
+__REF
_GLOBAL(create_hpte)
/* Convert linux-style PTE (r5) to low word of PPC-style PTE (r8) */
rlwinm r8,r5,32-9,30,30/* _PAGE_RW -> PP msb */
@@ -476,6 +478,7 @@ END_FTR_SECTION_IFCLR(CPU_FTR_NEED_COHERENT)
sync/* make sure pte updates get to memory */
blr
+ .previous
_ASM_NOKPROBE_SYMBOL(create_hpte)
.section .bss
@@ -496,6 +499,7 @@ htab_hash_searches:
*
* We assume that there is a hash table in use (Hash != 0).
*/
+__REF
_GLOBAL(flush_hash_pages)
/*
* We disable interrupts here, even on UP, because we want
@@ -632,6 +636,7 @@ _GLOBAL(flush_hash_pages)
SYNC_601
isync
blr
+ .previous
EXPORT_SYMBOL(flush_hash_pages)
_ASM_NOKPROBE_SYMBOL(flush_hash_pages)
diff --git a/arch/powerpc/mm/book3s32/mmu.c b/arch/powerpc/mm/book3s32/mmu.c
index d426eaf76bb0..3cf1177738ea 100644
--- a/arch/powerpc/mm/book3s32/mmu.c
+++ b/arch/powerpc/mm/book3s32/mmu.c
@@ -31,6 +31,8 @@
#include
+u8 __initdata early_hash[SZ_256K] __aligned(SZ_256K) = {0};
+
struct hash_pte *Hash;
static unsigned long Hash_size, Hash_mask;
unsigned long _SDR1;
@@ -425,15 +427,6 @@ void __init MMU_init_hw(void)
hash_mb2 = hash_mb = 32 - LG_HPTEG_SIZE - lg_n_hpteg;
if (lg_n_hpteg > 16)
hash_mb2 = 16 - LG_HPTEG_SIZE;
-
- /*
-* When KASAN is selected, there is already an early temporary hash
-* table and the switch to the final hash table is done later.
-*/
- if (IS_ENABLED(CONFIG_KASAN))
- return;
-
- MMU_init_hw_patch();
}
void __init MMU_init_hw_patch(void)
@@ -441,6 +434,9 @@ void __init
