Use and mention the predefined LINUX_GATEWAY_SPACE constant in the various important code sections which deal with the gateway page. Signed-off-by: Helge Deller <deller@gmx.de>
Remove two instruction from the hot path. The temporary move to %r9 is unneccessary, and the zero-inialization of pte happens twice. Signed-off-by: Helge Deller <deller@gmx.de>
This patch adds the necessary code to patch a running kernel at runtime to improve performance. The current implementation offers a few optimizations variants: - When running a SMP kernel on a single UP processor, unwanted assembler statements like locking functions are overwritten with NOPs. When multiple instructions shall be skipped, one branch instruction is used instead of multiple nop instructions. - In the UP case, some pdtlb and pitlb instructions are patched to become pdtlb,l and pitlb,l which only flushes the CPU-local tlb entries instead of broadcasting the flush to other CPUs in the system and thus may improve performance. - fic and fdc instructions are skipped if no I- or D-caches are installed. This should speed up qemu emulation and cacheless systems. - If no cache coherence is needed for IO operations, the relevant fdc and sync instructions in the sba and ccio drivers are replaced by nops. - On systems which share I- and D-TLBs and thus don't have a seperate instruction TLB, the pitlb instruction is replaced by a nop. Live-patching is done early in the boot process, just after having run the system inventory. No drivers are running and thus no external interrupts should arrive. So the hope is that no TLB exceptions will occur during the patching. If this turns out to be wrong we will probably need to do the patching in real-mode. Signed-off-by: Helge Deller <deller@gmx.de>
Helge noticed that the address of the os_hpmc handler was not being
correctly calculated in the hpmc macro. As a result, PDCE_CHECK would
fail to call os_hpmc:
<Cpu2> e800009802e00000 0000000000000000 CC_ERR_CHECK_HPMC
<Cpu2> 37000f7302e00000 8040004000000000 CC_ERR_CPU_CHECK_SUMMARY
<Cpu2> f600105e02e00000 fffffff0f0c00000 CC_MC_HPMC_MONARCH_SELECTED
<Cpu2> 140003b202e00000 000000000000000b CC_ERR_HPMC_STATE_ENTRY
<Cpu2> 5600100b02e00000 00000000000001a0 CC_MC_OS_HPMC_LEN_ERR
<Cpu2> 5600106402e00000 fffffff0f0438e70 CC_MC_BR_TO_OS_HPMC_FAILED
<Cpu2> e800009802e00000 0000000000000000 CC_ERR_CHECK_HPMC
<Cpu2> 37000f7302e00000 8040004000000000 CC_ERR_CPU_CHECK_SUMMARY
<Cpu2> 4000109f02e00000 0000000000000000 CC_MC_HPMC_INITIATED
<Cpu2> 4000101902e00000 0000000000000000 CC_MC_MULTIPLE_HPMCS
<Cpu2> 030010d502e00000 0000000000000000 CC_CPU_STOP
The address problem can be seen by dumping the fault vector:
0000000040159000 <fault_vector_20>:
40159000: 63 6f 77 73 stb r15,-2447(dp)
40159004: 20 63 61 6e ldil L%b747000,r3
40159008: 20 66 6c 79 ldil L%-1c3b3000,r3
...
40159020: 08 00 02 40 nop
40159024: 20 6e 60 02 ldil L%15d000,r3
40159028: 34 63 00 00 ldo 0(r3),r3
4015902c: e8 60 c0 02 bv,n r0(r3)
40159030: 08 00 02 40 nop
40159034: 00 00 00 00 break 0,0
40159038: c0 00 70 00 bb,*< r0,sar,40159840 <fault_vector_20+0x840>
4015903c: 00 00 00 00 break 0,0
Location 40159038 should contain the physical address of os_hpmc:
000000004015d000 <os_hpmc>:
4015d000: 08 1a 02 43 copy r26,r3
4015d004: 01 c0 08 a4 mfctl iva,r4
4015d008: 48 85 00 68 ldw 34(r4),r5
This patch moves the address setup into initialize_ivt to resolve the
above problem. I tested the change by dumping the HPMC entry after setup:
0000000040209020: 8000240
0000000040209024: 206a2004
0000000040209028: 34630ac0
000000004020902c: e860c002
0000000040209030: 8000240
0000000040209034: 1bdddce6
0000000040209038: 15d000
000000004020903c: 1a0
Signed-off-by: John David Anglin <dave.anglin@bell.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Helge Deller <deller@gmx.de>
This patch may resolve some races in TLB handling. Hopefully, TLB inserts are accesses and protected by spin lock. If not, we may need to IPI calls and do local purges on PA 2.0. Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
This change removes the PTE load and present check from the L2_ptep macro. The load and check for kernel pages is now done in the tlb_lock macro. This avoids a double load and check for user pages. The load and check for user pages is now done inside the lock so the fault handler can't be called while the entry is being updated. This version uses an ordered store to release the lock when the page table entry isn't present. It also corrects the check in the non SMP case. Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: Helge Deller <deller@gmx.de>
Fixes: 5b00ca0b8035 ("parisc: Restore possibility to execute 64-bit applications")
Signed-off-by: Helge Deller <deller@gmx.de>
Commit c8921d72e390 ("parisc: Fix and improve kernel stack unwinding")
broke booting of 64-bit kernels. On 64-bit kernels function pointers are
actually function descriptors which require dereferencing. In this patch
we instead declare functions in assembly code which are referenced from
C-code as external data pointers with the ENTRY() macro and thus can use
a simple external reference to the functions.
Signed-off-by: Helge Deller <deller@gmx.de>
Fixes: c8921d72e390 ("parisc: Fix and improve kernel stack unwinding")
Executing 64-bit applications was broken. This patch restores this support and cleans up some code paths. Signed-off-by: Helge Deller <deller@gmx.de>
This patchset fixes and improves stack unwinding a lot: 1. Show backward stack traces with up to 30 callsites 2. Add callinfo to ENTRY_CFI() such that every assembler function will get an entry in the unwind table 3. Use constants instead of numbers in call_on_stack() 4. Do not depend on CONFIG_KALLSYMS to generate backtraces. 5. Speed up backtrace generation Make sure you have this patch to GNU as installed: https://sourceware.org/ml/binutils/2018-07/msg00474.html Without this patch, unwind info in the kernel is often wrong for various functions. Signed-off-by: Helge Deller <deller@gmx.de>