- 08 Sep, 2020 11 commits
-
-
Michael Ellerman authored
The DSCR tests fail on systems that don't have DSCR, so check for the DSCR in hwcap and skip if it's not present. Signed-off-by:
Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20200819015727.1977134-5-mpe@ellerman.id.au
-
Michael Ellerman authored
utils.h provides have_hwcap() and have_hwcap2() which check for a feature bit. Those bits are defined in asm/cputable.h, so include it in utils.h so users of utils.h don't have to do it manually. Signed-off-by:
-
Michael Ellerman authored
This version of set_dscr() was added for the RFI flush test, and is fairly specific to it. It also clashes with the version of set_dscr() in dscr/dscr.h. So move it into the RFI flush test where it's used. Signed-off-by:
-
Michael Ellerman authored
On older systems this test takes longer to run (duh), give it five minutes which is long enough on a G5 970FX @ 1.6GHz. Signed-off-by:
-
Michael Ellerman authored
These platforms don't show the MMU in /proc/cpuinfo, but they always use hash, so teach using_hash_mmu() that. Signed-off-by:
-
Michael Ellerman authored
This test creates some threads, which write to TM SPRs, and then makes sure the registers maintain the correct values across context switches and contention with other threads. But currently the test finishes almost instantaneously, which reduces the chance of it hitting an interesting condition. So increase the number of loops, so it runs a bit longer, though still less than 2s on a Power8. Signed-off-by:
-
Michael Ellerman authored
This test tries to set affinity to CPUs that don't exist, especially if the set of online CPUs doesn't start at 0. But there's no real reason for it to use setaffinity in the first place, it's just trying to create lots of threads to cause contention. So drop the setaffinity entirely. Signed-off-by:
-
Michael Ellerman authored
Several of the TM tests fail spuriously if CPU 0 is offline, because they blindly try to affinitise to CPU 0. Fix them by picking any online CPU and using that instead. Signed-off-by:
-
Oliver O'Halloran authored
These annoy me every time I see them. Why are they here? They're not even needed for 80cols compliance. Signed-off-by:
Oliver O'Halloran <oohall@gmail.com> Signed-off-by:
-
Christophe Leroy authored
ftrace_graph_ret_addr() is always defined and returns 'ip' when CONFIG_FUNCTION GRAPH_TRACER is not set. So the #ifdef is not needed, remove it. Signed-off-by:
Christophe Leroy <christophe.leroy@csgroup.eu> Acked-by:
Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Signed-off-by:
-
Christophe Leroy authored
Enable pre-update addressing mode in __get_user_asm() and __put_user_asm() Signed-off-by:
Segher Boessenkool <segher@kernel.crashing.org> Signed-off-by:
-
- 02 Sep, 2020 27 commits
-
-
Christophe Leroy authored
At the time being, __put_user()/__get_user() and friends only use D-form addressing, with 0 offset. Ex: lwz reg1, 0(reg2) Give the compiler the opportunity to use other adressing modes whenever possible, to get more optimised code. Hereunder is a small exemple: struct test { u32 item1; u16 item2; u8 item3; u64 item4; }; int set_test_user(struct test __user *from, struct test __user *to) { int err; u32 item1; u16 item2; u8 item3; u64 item4; err = __get_user(item1, &from->item1); err |= __get_user(item2, &from->item2); err |= __get_user(item3, &from->item3); err |= __get_user(item4, &from->item4); err |= __put_user(item1, &to->item1); err |= __put_user(item2, &to->item2); err |= __put_user(item3, &to->item3); err |= __put_user(item4, &to->item4); return err; } Before the patch: 00000df0 <set_test_user>: df0: 94 21 ff f0 stwu r1,-16(r1) df4: 39 40 00 00 li r10,0 df8: 93 c1 00 08 stw r30,8(r1) dfc: 93 e1 00 0c stw r31,12(r1) e00: 7d 49 53 78 mr r9,r10 e04: 80 a3 00 00 lwz r5,0(r3) e08: 38 e3 00 04 addi r7,r3,4 e0c: 7d 46 53 78 mr r6,r10 e10: a0 e7 00 00 lhz r7,0(r7) e14: 7d 29 33 78 or r9,r9,r6 e18: 39 03 00 06 addi r8,r3,6 e1c: 7d 46 53 78 mr r6,r10 e20: 89 08 00 00 lbz r8,0(r8) e24: 7d 29 33 78 or r9,r9,r6 e28: 38 63 00 08 addi r3,r3,8 e2c: 7d 46 53 78 mr r6,r10 e30: 83 c3 00 00 lwz r30,0(r3) e34: 83 e3 00 04 lwz r31,4(r3) e38: 7d 29 33 78 or r9,r9,r6 e3c: 7d 43 53 78 mr r3,r10 e40: 90 a4 00 00 stw r5,0(r4) e44: 7d 29 1b 78 or r9,r9,r3 e48: 38 c4 00 04 addi r6,r4,4 e4c: 7d 43 53 78 mr r3,r10 e50: b0 e6 00 00 sth r7,0(r6) e54: 7d 29 1b 78 or r9,r9,r3 e58: 38 e4 00 06 addi r7,r4,6 e5c: 7d 43 53 78 mr r3,r10 e60: 99 07 00 00 stb r8,0(r7) e64: 7d 23 1b 78 or r3,r9,r3 e68: 38 84 00 08 addi r4,r4,8 e6c: 93 c4 00 00 stw r30,0(r4) e70: 93 e4 00 04 stw r31,4(r4) e74: 7c 63 53 78 or r3,r3,r10 e78: 83 c1 00 08 lwz r30,8(r1) e7c: 83 e1 00 0c lwz r31,12(r1) e80: 38 21 00 10 addi r1,r1,16 e84: 4e 80 00 20 blr After the patch: 00000dbc <set_test_user>: dbc: 39 40 00 00 li r10,0 dc0: 7d 49 53 78 mr r9,r10 dc4: 80 03 00 00 lwz r0,0(r3) dc8: 7d 48 53 78 mr r8,r10 dcc: a1 63 00 04 lhz r11,4(r3) dd0: 7d 29 43 78 or r9,r9,r8 dd4: 7d 48 53 78 mr r8,r10 dd8: 88 a3 00 06 lbz r5,6(r3) ddc: 7d 29 43 78 or r9,r9,r8 de0: 7d 48 53 78 mr r8,r10 de4: 80 c3 00 08 lwz r6,8(r3) de8: 80 e3 00 0c lwz r7,12(r3) dec: 7d 29 43 78 or r9,r9,r8 df0: 7d 43 53 78 mr r3,r10 df4: 90 04 00 00 stw r0,0(r4) df8: 7d 29 1b 78 or r9,r9,r3 dfc: 7d 43 53 78 mr r3,r10 e00: b1 64 00 04 sth r11,4(r4) e04: 7d 29 1b 78 or r9,r9,r3 e08: 7d 43 53 78 mr r3,r10 e0c: 98 a4 00 06 stb r5,6(r4) e10: 7d 23 1b 78 or r3,r9,r3 e14: 90 c4 00 08 stw r6,8(r4) e18: 90 e4 00 0c stw r7,12(r4) e1c: 7c 63 53 78 or r3,r3,r10 e20: 4e 80 00 20 blr Signed-off-by: -
Christophe Leroy authored
flush_instruction_cache() is never used on 8xx, remove it. Signed-off-by:
-
Stephen Rothwell authored
Currently, using llvm-objtool, this script just silently succeeds without actually do the intended checking. So this updates it to work properly. Firstly, llvm-objdump does not add target symbol names to the end of branches in its asm output, so we have to drop the branch to __start_initialization_multiplatform using its address. Secondly, v9 and 10 specify branch targets as .+<offset>, so we convert those to actual addresses. Thirdly, v10 and 11 error out on a vmlinux if given the -R option complaining that it is "not a dynamic object". The -R does not make any difference to the asm output, so remove it. Lastly, v11 produces asm that is very similar to Gnu objtool (at least as far as branches are concerned), so no further changes are necessary to make it work. Signed-off-by:
Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
Stephen Rothwell authored
This is considerably faster then parsing the objdump asm output. It will also make the enabling of llvm-objdump a little easier. Signed-off-by:
-
Stephen Rothwell authored
If we can't find the address of __end_interrupts, then we still exit successfully as that is the current behaviour. Signed-off-by:
-
Stephen Rothwell authored
Signed-off-by:
-
Stephen Rothwell authored
Signed-off-by:
-
Stephen Rothwell authored
Also start using sed -E and make all the separate expressions into a single one with comments. Pull the stripping of condition registers back into the sed command. Signed-off-by:
-
Stephen Rothwell authored
We don't use the raw hex instruction dump, so elide it and adjust the following expressions. Also use \s instead of [[:space:]] everywhere. Signed-off-by:
-
Stephen Rothwell authored
Also some minor style changes. There should still be no change in behaviour. Signed-off-by:
-
Stephen Rothwell authored
No functional change Signed-off-by:
-
Scott Cheloha authored
At memory hot-remove time we can retrieve an LMB's nid from its corresponding memory_block. There is no need to store the nid in multiple locations. Note that lmb_to_memblock() uses find_memory_block() to get the corresponding memory_block. As find_memory_block() runs in sub-linear time this approach is negligibly slower than what we do at present. In exchange for this lookup at hot-remove time we no longer need to call memory_add_physaddr_to_nid() during drmem_init() for each LMB. On powerpc, memory_add_physaddr_to_nid() is a linear search, so this spares us an O(n^2) initialization during boot. On systems with many LMBs that initialization overhead is palpable and disruptive. For example, on a box with 249854 LMBs we're seeing drmem_init() take upwards of 30 seconds to complete: [ 53.721639] drmem: initializing drmem v2 [ 80.604346] watchdog: BUG: soft lockup - CPU#65 stuck for 23s! [swapper/0:1] [ 80.604377] Modules linked in: [ 80.604389] CPU: 65 PID: 1 Comm: swapper/0 Not tainted 5.6.0-rc2+ #4 [ 80.604397] NIP: c0000000000a4980 LR: c0000000000a4940 CTR: 0000000000000000 [ 80.604407] REGS: c0002dbff8493830 TRAP: 0901 Not tainted (5.6.0-rc2+) [ 80.604412] MSR: 8000000002009033 <SF,VEC,EE,ME,IR,DR,RI,LE> CR: 44000248 XER: 0000000d [ 80.604431] CFAR: c0000000000a4a38 IRQMASK: 0 [ 80.604431] GPR00: c0000000000a4940 c0002dbff8493ac0 c000000001904400 c0003cfffffede30 [ 80.604431] GPR04: 0000000000000000 c000000000f4095a 000000000000002f 0000000010000000 [ 80.604431] GPR08: c0000bf7ecdb7fb8 c0000bf7ecc2d3c8 0000000000000008 c00c0002fdfb2001 [ 80.604431] GPR12: 0000000000000000 c00000001e8ec200 [ 80.604477] NIP [c0000000000a4980] hot_add_scn_to_nid+0xa0/0x3e0 [ 80.604486] LR [c0000000000a4940] hot_add_scn_to_nid+0x60/0x3e0 [ 80.604492] Call Trace: [ 80.604498] [c0002dbff8493ac0] [c0000000000a4940] hot_add_scn_to_nid+0x60/0x3e0 (unreliable) [ 80.604509] [c0002dbff8493b20] [c000000000087c10] memory_add_physaddr_to_nid+0x20/0x60 [ 80.604521] [c0002dbff8493b40] [c0000000010d4880] drmem_init+0x25c/0x2f0 [ 80.604530] [c0002dbff8493c10] [c000000000010154] do_one_initcall+0x64/0x2c0 [ 80.604540] [c0002dbff8493ce0] [c0000000010c4aa0] kernel_init_freeable+0x2d8/0x3a0 [ 80.604550] [c0002dbff8493db0] [c000000000010824] kernel_init+0x2c/0x148 [ 80.604560] [c0002dbff8493e20] [c00000000000b648] ret_from_kernel_thread+0x5c/0x74 [ 80.604567] Instruction dump: [ 80.604574] 392918e8 e9490000 e90a000a e92a0000 80ea000c 1d080018 3908ffe8 7d094214 [ 80.604586] 7fa94040 419d00dc e9490010 714a0088 <2faa0008> 409e00ac e9490000 7fbe5040 [ 89.047390] drmem: 249854 LMB(s) With a patched kernel on the same machine we're no longer seeing the soft lockup. drmem_init() now completes in negligible time, even when the LMB count is large. Fixes: b2d3b5ee ("powerpc/pseries: Track LMB nid instead of using device tree") Signed-off-by:
Scott Cheloha <cheloha@linux.ibm.com> Reviewed-by:
Nathan Lynch <nathanl@linux.ibm.com> Signed-off-by:
-
Christophe Leroy authored
Nothing prevents flush_cache_instruction() from being writen in C. Do it to improve readability and maintainability. This function is only use by low level callers, it is not intended to be used by module. Don't export it. Signed-off-by:
-
Christophe Leroy authored
Nothing prevents flush_cache_instruction() from being writen in C. Do it to improve readability and maintainability. This function is very small and isn't called from assembly, make it static inline in asm/cacheflush.h Signed-off-by:
-
Christophe Leroy authored
flush_instruction_cache() belongs to the cache flushing function family. Move its prototype in asm/cacheflush.h Signed-off-by:
-
Christophe Leroy authored
The only callers of flush_instruction_cache() are: arch/powerpc/kernel/swsusp_booke.S: bl flush_instruction_cache arch/powerpc/mm/nohash/40x.c: flush_instruction_cache(); arch/powerpc/mm/nohash/44x.c: flush_instruction_cache(); arch/powerpc/mm/nohash/fsl_booke.c: flush_instruction_cache(); arch/powerpc/platforms/44x/machine_check.c: flush_instruction_cache(); arch/powerpc/platforms/44x/machine_check.c: flush_instruction_cache(); This function is not used by book3s/32, drop it. Signed-off-by:
-
Nathan Lynch authored
The drmem lmb list can have hundreds of thousands of entries, and unfortunately lookups take the form of linear searches. As long as this is the case, traversals have the potential to monopolize the CPU and provoke lockup reports, workqueue stalls, and the like unless they explicitly yield. Rather than placing cond_resched() calls within various for_each_drmem_lmb() loop blocks in the code, put it in the iteration expression of the loop macro itself so users can't omit it. Introduce a drmem_lmb_next() iteration helper function which calls cond_resched() at a regular interval during array traversal. Each iteration of the loop in DLPAR code paths can involve around ten RTAS calls which can each take up to 250us, so this ensures the check is performed at worst every few milliseconds. Fixes: 6c6ea537 ("powerpc/mm: Separate ibm, dynamic-memory data from DT format") Signed-off-by:
-
Christophe Leroy authored
_nmask_and_or_msr() is only used at two places to set MSR_IP. The SYNC is unnecessary as the users are not PowerPC 601. Can be easily writen in C. Do it, and drop _nmask_and_or_msr() Signed-off-by:
-
Stephen Kitt authored
The i2c probe functions here don't use the id information provided in their second argument, so the single-parameter i2c probe function ("probe_new") can be used instead. This avoids scanning the identifier tables during probes. Signed-off-by:Stephen Kitt <steve@sk2.org> Acked-by:
Wolfram Sang <wsa@kernel.org> Reviewed-by:
Luca Ceresoli <luca@lucaceresoli.net> Signed-off-by:
-
Scott Cheloha authored
The H_GetPerformanceCounterInfo (GPCI) PHYP hypercall has a subcall, Affinity_Domain_Info_By_Partition, which returns, among other things, a "partition affinity score" for a given LPAR. This score, a value on [0-100], represents the processor-memory affinity for the LPAR in question. A score of 0 indicates the worst possible affinity while a score of 100 indicates perfect affinity. The score can be used to reason about performance. This patch adds the score for the local LPAR to the lparcfg procfile under a new 'partition_affinity_score' key. Signed-off-by:
Tyrel Datwyler <tyreld@linux.ibm.com> Acked-by:
-
Scott Cheloha authored
The H_GetPerformanceCounterInfo (GPCI) hypercall input/output structs are useful to modules outside of perf/, so move them into asm/hvcall.h to live alongside the other powerpc hypercall structs. Leave the perf-specific GPCI stuff in perf/hv-gpci.h. Signed-off-by:
-
Christophe Leroy authored
Those function have never existed. Drop their declaration. Signed-off-by:
-
Christophe Leroy authored
Those two functions have been unused since commit identified below. Drop them. Fixes: 31bfdb03 ("powerpc: Use instruction emulation infrastructure to handle alignment faults") Signed-off-by:
-
Christophe Leroy authored
Since the commit identified below, the forward declaration of struct irqaction is useless. Drop it. Fixes: b709c083 ("ppc64: move stack switching up in interrupt processing") Signed-off-by:
-
Christophe Leroy authored
Since commit identified below, the forward declaration of struct irq_chip is useless (was struct hw_interrupt_type at that time) Remove it, together with the associated comment. Fixes: c0ad90a3 ("[PATCH] genirq: add ->retrigger() irq op to consolidate hw_irq_resend()") Signed-off-by:
-
Colin Ian King authored
There are spelling mistakes in two debug messages. As recommended by Wolfram Sang, these can be removed as there is plenty of debug in the driver core. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Reviewed-by:
-
Christophe Leroy authored
The assembler says: arch/powerpc/kernel/head_32.S:1095: Warning: invalid register expression It's objecting to the use of r0 as the RA argument. That's because when RA = 0 the literal value 0 is used, rather than the content of r0, making the use of r0 in the source potentially confusing. Fix it to use a literal 0, the generated code is identical. Signed-off-by:
-
- 24 Aug, 2020 2 commits
-
-
Oliver O'Halloran authored
Building with W=1 results in the following warning: In file included from arch/powerpc/platforms/powernv/vas-fault.c:16: ./arch/powerpc/include/asm/icswx.h:159:1: error: alignment 1 of ‘struct coprocessor_request_block’ is less than 16 [-Werror=packed-not-aligned] 159 | } __packed; | ^ ./arch/powerpc/include/asm/icswx.h:159:1: error: alignment 1 of ‘struct coprocessor_request_block’ is less than 16 [-Werror=packed-not-aligned] ./arch/powerpc/include/asm/icswx.h:159:1: error: alignment 1 of ‘struct coprocessor_request_block’ is less than 16 [-Werror=packed-not-aligned] ./arch/powerpc/include/asm/icswx.h:159:1: error: alignment 1 of ‘struct coprocessor_request_block’ is less than 16 [-Werror=packed-not-aligned] cc1: all warnings being treated as errors This happens because coprocessor_request_block includes several sub-structures with an alignment specified using the __aligned(XX) attribute. The problem comes from coprocessor_request_block having the __packed attribute. Packing the structure causes the preferred alignment of the nested structures to be ignored and we get the warnings as a result. This isn't a problem in practice since the struct is defined with explicit padding in the form of reserved fields, but we'd like to get rid of the spurious warnings. The simplest solution is to remove the packed attribute and use a BUILD_BUG_ON() to ensure the struct is the correct (expected by HW) size compile time. Also add a __aligned(128) to the request block structure since Book4 for P8 suggests the HW requires it to be aligned to a 128 byte boundary. There's a similar requirement for P9 since the COPY and PASTE instructions used to invoke VAS/NX accelerators operates on a cache line boundary. Signed-off-by: -
Oliver O'Halloran authored
Comments opening with /** are parsed by kerneldoc and this causes the following warning to be printed: arch/powerpc/platforms/powernv/opal-prd.c:31: warning: cannot understand function prototype: 'struct opal_prd_msg_queue_item ' opal_prd_mesg_queue_item is an internal data structure so there's no real need for it to be documented at all. Fix up the comment to squash the warning. Signed-off-by:
Joel Stanley <joel@jms.id.au> Signed-off-by:
-
