| Message ID | 20180605161911.26434-1-lionel.g.landwerlin@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Quoting Lionel Landwerlin (2018-06-05 17:19:11) > We would like to make use of perf in Vulkan. The Vulkan API is much > lower level than OpenGL, with applications directly exposed to the > concept of command buffers (pretty much equivalent to our batch > buffers). In Vulkan, queries are always limited in scope to a command > buffer. In OpenGL, the lack of command buffer concept meant that > queries' duration could span multiple command buffers. > > With that restriction gone in Vulkan, we would like to simplify > measuring performance just by measuring the deltas between the counter > snapshots written by 2 MI_RECORD_PERF_COUNT commands, rather than the > more complex scheme with currently have in the GL driver, using 2 > MI_RECORD_PERF_COUNT commands and doing some post processing on the > stream of OA reports to remove any unrelated deltas in the OA stream. > > Disabling preemption only apply to the context with which want to > query performance counters and is considered a privileged operation > (disabled by default). and protected by CAP_SYS_ADMIN (give or take it requiring root privilege to let an ordinary user set it themselves). Please make that clear in the commit log for the likes of myself who panicked at the mere thought of being able to stop preemption (because it's an outright priority inversion). Note that this will not escape a fast reset timer so the batch had better be short, or they will end up being banned. > diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c > index 7d63c6d2f687..417ca93ea606 100644 > --- a/drivers/gpu/drm/i915/intel_lrc.c > +++ b/drivers/gpu/drm/i915/intel_lrc.c > @@ -2099,7 +2099,9 @@ static int gen8_emit_bb_start(struct i915_request *rq, > * > * That satisfies both the GPGPU w/a and our heavy-handed paranoia. > */ > - *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; > + *cs++ = MI_ARB_ON_OFF | I'd prefer having it as a u32 enable/disable. Could it be intel_context even, i.e. selectable on engine? *cs++ = MI_ARB_ON_OFF | rq->hw_context->arb_enable; -Chris
On 05/06/18 17:28, Chris Wilson wrote: > Quoting Lionel Landwerlin (2018-06-05 17:19:11) >> We would like to make use of perf in Vulkan. The Vulkan API is much >> lower level than OpenGL, with applications directly exposed to the >> concept of command buffers (pretty much equivalent to our batch >> buffers). In Vulkan, queries are always limited in scope to a command >> buffer. In OpenGL, the lack of command buffer concept meant that >> queries' duration could span multiple command buffers. >> >> With that restriction gone in Vulkan, we would like to simplify >> measuring performance just by measuring the deltas between the counter >> snapshots written by 2 MI_RECORD_PERF_COUNT commands, rather than the >> more complex scheme with currently have in the GL driver, using 2 >> MI_RECORD_PERF_COUNT commands and doing some post processing on the >> stream of OA reports to remove any unrelated deltas in the OA stream. >> >> Disabling preemption only apply to the context with which want to >> query performance counters and is considered a privileged operation >> (disabled by default). > and protected by CAP_SYS_ADMIN (give or take it requiring root > privilege to let an ordinary user set it themselves). > > Please make that clear in the commit log for the likes of myself who > panicked at the mere thought of being able to stop preemption (because > it's an outright priority inversion). Note that this will not escape a > fast reset timer so the batch had better be short, or they will end up > being banned. > Sure, I thought people would be scared. Adding it locally. >> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c >> index 7d63c6d2f687..417ca93ea606 100644 >> --- a/drivers/gpu/drm/i915/intel_lrc.c >> +++ b/drivers/gpu/drm/i915/intel_lrc.c >> @@ -2099,7 +2099,9 @@ static int gen8_emit_bb_start(struct i915_request *rq, >> * >> * That satisfies both the GPGPU w/a and our heavy-handed paranoia. >> */ >> - *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; >> + *cs++ = MI_ARB_ON_OFF | > I'd prefer having it as a u32 enable/disable. Could it be intel_context > even, i.e. selectable on engine? > > *cs++ = MI_ARB_ON_OFF | rq->hw_context->arb_enable; Yeah, I just thought about putting it on the engine. Will do that too. Right now it only make sense on the RCS anyway, that's where the OA unit works. Thanks, - Lionel > -Chris >
Hi Lionel,
Thank you for the patch! Yet something to improve:
[auto build test ERROR on drm-intel/for-linux-next]
[also build test ERROR on v4.17 next-20180605]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Lionel-Landwerlin/drm-i915-perf-allow-holding-preemption-on-filtered-ctx/20180607-005605
base: git://anongit.freedesktop.org/drm-intel for-linux-next
config: x86_64-randconfig-x011-201822 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64
All errors (new ones prefixed by >>):
drivers/gpu//drm/i915/intel_lrc.c: In function 'gen8_emit_bb_start':
>> drivers/gpu//drm/i915/intel_lrc.c:1981:6: error: 'struct i915_request' has no member named 'gem_context'
(rq->gem_context->perf_disabled_preemption ?
^~
vim +1981 drivers/gpu//drm/i915/intel_lrc.c
1934
1935 static int gen8_emit_bb_start(struct i915_request *rq,
1936 u64 offset, u32 len,
1937 const unsigned int flags)
1938 {
1939 u32 *cs;
1940 int ret;
1941
1942 /* Don't rely in hw updating PDPs, specially in lite-restore.
1943 * Ideally, we should set Force PD Restore in ctx descriptor,
1944 * but we can't. Force Restore would be a second option, but
1945 * it is unsafe in case of lite-restore (because the ctx is
1946 * not idle). PML4 is allocated during ppgtt init so this is
1947 * not needed in 48-bit.*/
1948 if (rq->ctx->ppgtt &&
1949 (intel_engine_flag(rq->engine) & rq->ctx->ppgtt->pd_dirty_rings) &&
1950 !i915_vm_is_48bit(&rq->ctx->ppgtt->base) &&
1951 !intel_vgpu_active(rq->i915)) {
1952 ret = intel_logical_ring_emit_pdps(rq);
1953 if (ret)
1954 return ret;
1955
1956 rq->ctx->ppgtt->pd_dirty_rings &= ~intel_engine_flag(rq->engine);
1957 }
1958
1959 cs = intel_ring_begin(rq, 6);
1960 if (IS_ERR(cs))
1961 return PTR_ERR(cs);
1962
1963 /*
1964 * WaDisableCtxRestoreArbitration:bdw,chv
1965 *
1966 * We don't need to perform MI_ARB_ENABLE as often as we do (in
1967 * particular all the gen that do not need the w/a at all!), if we
1968 * took care to make sure that on every switch into this context
1969 * (both ordinary and for preemption) that arbitrartion was enabled
1970 * we would be fine. However, there doesn't seem to be a downside to
1971 * being paranoid and making sure it is set before each batch and
1972 * every context-switch.
1973 *
1974 * Note that if we fail to enable arbitration before the request
1975 * is complete, then we do not see the context-switch interrupt and
1976 * the engine hangs (with RING_HEAD == RING_TAIL).
1977 *
1978 * That satisfies both the GPGPU w/a and our heavy-handed paranoia.
1979 */
1980 *cs++ = MI_ARB_ON_OFF |
> 1981 (rq->gem_context->perf_disabled_preemption ?
1982 MI_ARB_DISABLE : MI_ARB_ENABLE);
1983
1984 /* FIXME(BDW): Address space and security selectors. */
1985 *cs++ = MI_BATCH_BUFFER_START_GEN8 |
1986 (flags & I915_DISPATCH_SECURE ? 0 : BIT(8)) |
1987 (flags & I915_DISPATCH_RS ? MI_BATCH_RESOURCE_STREAMER : 0);
1988 *cs++ = lower_32_bits(offset);
1989 *cs++ = upper_32_bits(offset);
1990
1991 *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
1992 *cs++ = MI_NOOP;
1993 intel_ring_advance(rq, cs);
1994
1995 return 0;
1996 }
1997
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
Hi Lionel,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on drm-intel/for-linux-next]
[also build test WARNING on v4.17 next-20180605]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Lionel-Landwerlin/drm-i915-perf-allow-holding-preemption-on-filtered-ctx/20180607-005605
base: git://anongit.freedesktop.org/drm-intel for-linux-next
reproduce: make htmldocs
All warnings (new ones prefixed by >>):
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
include/net/mac80211.h:2083: warning: bad line: >
net/mac80211/sta_info.h:586: warning: Function parameter or member 'rx_stats_avg' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'rx_stats_avg.signal' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'rx_stats_avg.chain_signal' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.filtered' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.retry_failed' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.retry_count' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.lost_packets' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.last_tdls_pkt_time' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.msdu_retries' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.msdu_failed' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.last_ack' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.last_ack_signal' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'status_stats.ack_signal_filled' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'tx_stats.packets' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'tx_stats.bytes' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'tx_stats.last_rate' not described in 'sta_info'
net/mac80211/sta_info.h:586: warning: Function parameter or member 'tx_stats.msdu' not described in 'sta_info'
kernel/sched/fair.c:3731: warning: Function parameter or member 'flags' not described in 'attach_entity_load_avg'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_excl.cb' not described in 'dma_buf'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_excl.poll' not described in 'dma_buf'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_excl.active' not described in 'dma_buf'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_shared.cb' not described in 'dma_buf'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_shared.poll' not described in 'dma_buf'
include/linux/dma-buf.h:307: warning: Function parameter or member 'cb_shared.active' not described in 'dma_buf'
include/linux/dma-fence-array.h:54: warning: Function parameter or member 'work' not described in 'dma_fence_array'
include/linux/gpio/driver.h:142: warning: Function parameter or member 'request_key' not described in 'gpio_irq_chip'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.sign' not described in 'iio_chan_spec'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.realbits' not described in 'iio_chan_spec'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.storagebits' not described in 'iio_chan_spec'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.shift' not described in 'iio_chan_spec'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.repeat' not described in 'iio_chan_spec'
include/linux/iio/iio.h:270: warning: Function parameter or member 'scan_type.endianness' not described in 'iio_chan_spec'
include/linux/iio/hw-consumer.h:1: warning: no structured comments found
include/linux/input/sparse-keymap.h:46: warning: Function parameter or member 'sw' not described in 'key_entry'
include/linux/mtd/rawnand.h:752: warning: Function parameter or member 'timings.sdr' not described in 'nand_data_interface'
include/linux/mtd/rawnand.h:817: warning: Function parameter or member 'buf' not described in 'nand_op_data_instr'
include/linux/mtd/rawnand.h:817: warning: Function parameter or member 'buf.in' not described in 'nand_op_data_instr'
include/linux/mtd/rawnand.h:817: warning: Function parameter or member 'buf.out' not described in 'nand_op_data_instr'
include/linux/mtd/rawnand.h:863: warning: Function parameter or member 'ctx' not described in 'nand_op_instr'
include/linux/mtd/rawnand.h:863: warning: Function parameter or member 'ctx.cmd' not described in 'nand_op_instr'
include/linux/mtd/rawnand.h:863: warning: Function parameter or member 'ctx.addr' not described in 'nand_op_instr'
include/linux/mtd/rawnand.h:863: warning: Function parameter or member 'ctx.data' not described in 'nand_op_instr'
include/linux/mtd/rawnand.h:863: warning: Function parameter or member 'ctx.waitrdy' not described in 'nand_op_instr'
include/linux/mtd/rawnand.h:1010: warning: Function parameter or member 'ctx' not described in 'nand_op_parser_pattern_elem'
include/linux/mtd/rawnand.h:1010: warning: Function parameter or member 'ctx.addr' not described in 'nand_op_parser_pattern_elem'
include/linux/mtd/rawnand.h:1010: warning: Function parameter or member 'ctx.data' not described in 'nand_op_parser_pattern_elem'
include/linux/mtd/rawnand.h:1313: warning: Function parameter or member 'manufacturer.desc' not described in 'nand_chip'
include/linux/mtd/rawnand.h:1313: warning: Function parameter or member 'manufacturer.priv' not described in 'nand_chip'
include/linux/regulator/driver.h:222: warning: Function parameter or member 'resume_early' not described in 'regulator_ops'
drivers/regulator/core.c:4306: warning: Excess function parameter 'state' description in 'regulator_suspend_late'
arch/s390/include/asm/cio.h:245: warning: Function parameter or member 'esw.esw0' not described in 'irb'
arch/s390/include/asm/cio.h:245: warning: Function parameter or member 'esw.esw1' not described in 'irb'
arch/s390/include/asm/cio.h:245: warning: Function parameter or member 'esw.esw2' not described in 'irb'
arch/s390/include/asm/cio.h:245: warning: Function parameter or member 'esw.esw3' not described in 'irb'
arch/s390/include/asm/cio.h:245: warning: Function parameter or member 'esw.eadm' not described in 'irb'
drivers/usb/typec/mux.c:186: warning: Function parameter or member 'mux' not described in 'typec_mux_unregister'
drivers/usb/typec/mux.c:186: warning: Excess function parameter 'sw' description in 'typec_mux_unregister'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_pin' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_unpin' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_res_obj' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_get_sg_table' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_import_sg_table' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_vmap' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_vunmap' not described in 'drm_driver'
include/drm/drm_drv.h:610: warning: Function parameter or member 'gem_prime_mmap' not described in 'drm_driver'
drivers/gpu/drm/i915/i915_vma.h:48: warning: cannot understand function prototype: 'struct i915_vma '
drivers/gpu/drm/i915/i915_vma.h:1: warning: no structured comments found
>> drivers/gpu/drm/i915/i915_perf.c:365: warning: Function parameter or member 'context_disable_preemption' not described in 'perf_open_properties'
include/drm/tinydrm/tinydrm.h:34: warning: Function parameter or member 'fb_dirty' not described in 'tinydrm_device'
drivers/gpu/drm/tinydrm/mipi-dbi.c:272: warning: Function parameter or member 'crtc_state' not described in 'mipi_dbi_enable_flush'
drivers/gpu/drm/tinydrm/mipi-dbi.c:272: warning: Function parameter or member 'plane_state' not described in 'mipi_dbi_enable_flush'
include/media/v4l2-dev.h:42: warning: Enum value 'VFL_TYPE_MAX' not described in enum 'vfl_devnode_type'
include/linux/skbuff.h:850: warning: Function parameter or member 'dev_scratch' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'ip_defrag_offset' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'skb_mstamp' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member '__cloned_offset' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'head_frag' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member '__unused' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member '__pkt_type_offset' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'pfmemalloc' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'encapsulation' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'encap_hdr_csum' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'csum_valid' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'csum_complete_sw' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'csum_level' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'inner_protocol_type' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'remcsum_offload' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'offload_fwd_mark' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'offload_mr_fwd_mark' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'sender_cpu' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'reserved_tailroom' not described in 'sk_buff'
include/linux/skbuff.h:850: warning: Function parameter or member 'inner_ipproto' not described in 'sk_buff'
include/net/sock.h:234: warning: Function parameter or member 'skc_addrpair' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_portpair' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_ipv6only' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_net_refcnt' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_v6_daddr' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_v6_rcv_saddr' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_cookie' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_listener' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_tw_dr' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_rcv_wnd' not described in 'sock_common'
include/net/sock.h:234: warning: Function parameter or member 'skc_tw_rcv_nxt' not described in 'sock_common'
include/net/sock.h:488: warning: Function parameter or member 'sk_backlog.rmem_alloc' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'sk_backlog.len' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'sk_backlog.head' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'sk_backlog.tail' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'sk_wq_raw' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'tcp_rtx_queue' not described in 'sock'
include/net/sock.h:488: warning: Function parameter or member 'sk_route_forced_caps' not described in 'sock'
include/linux/netdevice.h:1955: warning: Function parameter or member 'adj_list.upper' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'adj_list.lower' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'gso_partial_features' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'switchdev_ops' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'l3mdev_ops' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'xfrmdev_ops' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'name_assign_type' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'ieee802154_ptr' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'mpls_ptr' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'xdp_prog' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'gro_flush_timeout' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'nf_hooks_ingress' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member '____cacheline_aligned_in_smp' not described in 'net_device'
include/linux/netdevice.h:1955: warning: Function parameter or member 'qdisc_hash' not described in 'net_device'
include/linux/phylink.h:56: warning: Function parameter or member '__ETHTOOL_DECLARE_LINK_MODE_MASK(advertising' not described in 'phylink_link_state'
include/linux/phylink.h:56: warning: Function parameter or member '__ETHTOOL_DECLARE_LINK_MODE_MASK(lp_advertising' not described in 'phylink_link_state'
include/linux/rcupdate.h:572: ERROR: Unexpected indentation.
include/linux/rcupdate.h:576: ERROR: Unexpected indentation.
include/linux/rcupdate.h:580: WARNING: Block quote ends without a blank line; unexpected unindent.
include/linux/rcupdate.h:582: WARNING: Block quote ends without a blank line; unexpected unindent.
include/linux/rcupdate.h:582: WARNING: Inline literal start-string without end-string.
Documentation/crypto/crypto_engine.rst:13: ERROR: Unexpected indentation.
Documentation/crypto/crypto_engine.rst:15: WARNING: Block quote ends without a blank line; unexpected unindent.
kernel/time/timer.c:1259: ERROR: Unexpected indentation.
kernel/time/timer.c:1261: ERROR: Unexpected indentation.
kernel/time/timer.c:1262: WARNING: Block quote ends without a blank line; unexpected unindent.
include/linux/wait.h:110: WARNING: Block quote ends without a blank line; unexpected unindent.
include/linux/wait.h:113: ERROR: Unexpected indentation.
include/linux/wait.h:115: WARNING: Block quote ends without a blank line; unexpected unindent.
kernel/time/hrtimer.c:1129: WARNING: Block quote ends without a blank line; unexpected unindent.
kernel/signal.c:327: WARNING: Inline literal start-string without end-string.
Documentation/driver-api/device_connection.rst:42: ERROR: Error in "kernel-doc" directive:
maximum 4 argument(s) allowed, 7 supplied.
vim +365 drivers/gpu/drm/i915/i915_perf.c
eec688e1 Robert Bragg 2016-11-07 @365
:::::: The code at line 365 was first introduced by commit
:::::: eec688e1420da584afb36ffa5f0cad75f53cf286 drm/i915: Add i915 perf infrastructure
:::::: TO: Robert Bragg <robert@sixbynine.org>
:::::: CC: Daniel Vetter <daniel.vetter@ffwll.ch>
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
diff --git a/drivers/gpu/drm/i915/i915_gem_context.h b/drivers/gpu/drm/i915/i915_gem_context.h index f40d85448a28..17e8efe0397e 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.h +++ b/drivers/gpu/drm/i915/i915_gem_context.h @@ -187,6 +187,12 @@ struct i915_gem_context { /** remap_slice: Bitmask of cache lines that need remapping */ u8 remap_slice; + /* + * perf_disabled_preemption: Disable preemption for batch buffers + * submitted on this context. + */ + bool perf_disabled_preemption; + /** handles_vma: rbtree to look up our context specific obj/vma for * the user handle. (user handles are per fd, but the binding is * per vm, which may be one per context or shared with the global GTT) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 02f8034bb699..735babf272a9 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -354,6 +354,7 @@ struct perf_open_properties { u32 sample_flags; u64 single_context:1; + u64 context_disable_preemption:1; u64 ctx_handle; /* OA sampling state */ @@ -1362,6 +1363,8 @@ static void i915_oa_stream_destroy(struct i915_perf_stream *stream) mutex_lock(&dev_priv->drm.struct_mutex); dev_priv->perf.oa.exclusive_stream = NULL; dev_priv->perf.oa.ops.disable_metric_set(dev_priv); + if (stream->ctx) + stream->ctx->perf_disabled_preemption = false; mutex_unlock(&dev_priv->drm.struct_mutex); free_oa_buffer(dev_priv); @@ -2200,6 +2203,9 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream, goto err_enable; } + if (props->context_disable_preemption) + stream->ctx->perf_disabled_preemption = true; + stream->ops = &i915_oa_stream_ops; dev_priv->perf.oa.exclusive_stream = stream; @@ -2663,6 +2669,14 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, } } + if (props->context_disable_preemption) { + if (!specific_ctx) { + ret = -EINVAL; + goto err; + } + privileged_op = true; + } + /* * On Haswell the OA unit supports clock gating off for a specific * context and in this mode there's no visibility of metrics for the @@ -2677,8 +2691,10 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, * MI_REPORT_PERF_COUNT commands and so consider it a privileged op to * enable the OA unit by default. */ - if (IS_HASWELL(dev_priv) && specific_ctx) + if (IS_HASWELL(dev_priv) && specific_ctx && + !props->context_disable_preemption) { privileged_op = false; + } /* Similar to perf's kernel.perf_paranoid_cpu sysctl option * we check a dev.i915.perf_stream_paranoid sysctl option @@ -2687,7 +2703,7 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, */ if (privileged_op && i915_perf_stream_paranoid && !capable(CAP_SYS_ADMIN)) { - DRM_DEBUG("Insufficient privileges to open system-wide i915 perf stream\n"); + DRM_DEBUG("Insufficient privileges to open i915 perf stream\n"); ret = -EACCES; goto err_ctx; } @@ -2879,6 +2895,10 @@ static int read_properties_unlocked(struct drm_i915_private *dev_priv, props->oa_periodic = true; props->oa_period_exponent = value; break; + case DRM_I915_PERF_PROP_DISABLE_PREEMPTION: + if (value) + props->context_disable_preemption = 1; + break; case DRM_I915_PERF_PROP_MAX: MISSING_CASE(id); return -EINVAL; diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 7d63c6d2f687..417ca93ea606 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -2099,7 +2099,9 @@ static int gen8_emit_bb_start(struct i915_request *rq, * * That satisfies both the GPGPU w/a and our heavy-handed paranoia. */ - *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; + *cs++ = MI_ARB_ON_OFF | + (rq->gem_context->perf_disabled_preemption ? + MI_ARB_DISABLE : MI_ARB_ENABLE); /* FIXME(BDW): Address space and security selectors. */ *cs++ = MI_BATCH_BUFFER_START_GEN8 | diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 3d9c44f35df4..4cb80082df7e 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -1553,6 +1553,14 @@ enum drm_i915_perf_property_id { */ DRM_I915_PERF_PROP_OA_EXPONENT, + /** + * Specifying this property is only valid when specify a context to + * filter with DRM_I915_PERF_PROP_CTX_HANDLE. Specifying this property + * will disable preemption for the particular context we want to + * measure. + */ + DRM_I915_PERF_PROP_DISABLE_PREEMPTION, + DRM_I915_PERF_PROP_MAX /* non-ABI */ };
We would like to make use of perf in Vulkan. The Vulkan API is much lower level than OpenGL, with applications directly exposed to the concept of command buffers (pretty much equivalent to our batch buffers). In Vulkan, queries are always limited in scope to a command buffer. In OpenGL, the lack of command buffer concept meant that queries' duration could span multiple command buffers. With that restriction gone in Vulkan, we would like to simplify measuring performance just by measuring the deltas between the counter snapshots written by 2 MI_RECORD_PERF_COUNT commands, rather than the more complex scheme with currently have in the GL driver, using 2 MI_RECORD_PERF_COUNT commands and doing some post processing on the stream of OA reports to remove any unrelated deltas in the OA stream. Disabling preemption only apply to the context with which want to query performance counters and is considered a privileged operation (disabled by default). Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> --- drivers/gpu/drm/i915/i915_gem_context.h | 6 ++++++ drivers/gpu/drm/i915/i915_perf.c | 24 ++++++++++++++++++++++-- drivers/gpu/drm/i915/intel_lrc.c | 4 +++- include/uapi/drm/i915_drm.h | 8 ++++++++ 4 files changed, 39 insertions(+), 3 deletions(-)