public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	stable@vger.kernel.org, Stephane Eranian <eranian@google.com>,
	"Peter Zijlstra (Intel)" <peterz@infradead.org>,
	Kan Liang <kan.liang@linux.intel.com>
Subject: [PATCH 5.4 47/77] perf/x86/intel/uncore: Fix broken read_counter() for SNB IMC PMU
Date: Fri,  2 Sep 2022 14:18:56 +0200	[thread overview]
Message-ID: <20220902121405.211947115@linuxfoundation.org> (raw)
In-Reply-To: <20220902121403.569927325@linuxfoundation.org>

From: Stephane Eranian <eranian@google.com>

commit 11745ecfe8fea4b4a4c322967a7605d2ecbd5080 upstream.

Existing code was generating bogus counts for the SNB IMC bandwidth counters:

$ perf stat -a -I 1000 -e uncore_imc/data_reads/,uncore_imc/data_writes/
     1.000327813           1,024.03 MiB  uncore_imc/data_reads/
     1.000327813              20.73 MiB  uncore_imc/data_writes/
     2.000580153         261,120.00 MiB  uncore_imc/data_reads/
     2.000580153              23.28 MiB  uncore_imc/data_writes/

The problem was introduced by commit:
  07ce734dd8ad ("perf/x86/intel/uncore: Clean up client IMC")

Where the read_counter callback was replace to point to the generic
uncore_mmio_read_counter() function.

The SNB IMC counters are freerunnig 32-bit counters laid out contiguously in
MMIO. But uncore_mmio_read_counter() is using a readq() call to read from
MMIO therefore reading 64-bit from MMIO. Although this is okay for the
uncore_perf_event_update() function because it is shifting the value based
on the actual counter width to compute a delta, it is not okay for the
uncore_pmu_event_start() which is simply reading the counter  and therefore
priming the event->prev_count with a bogus value which is responsible for
causing bogus deltas in the perf stat command above.

The fix is to reintroduce the custom callback for read_counter for the SNB
IMC PMU and use readl() instead of readq(). With the change the output of
perf stat is back to normal:
$ perf stat -a -I 1000 -e uncore_imc/data_reads/,uncore_imc/data_writes/
     1.000120987             296.94 MiB  uncore_imc/data_reads/
     1.000120987             138.42 MiB  uncore_imc/data_writes/
     2.000403144             175.91 MiB  uncore_imc/data_reads/
     2.000403144              68.50 MiB  uncore_imc/data_writes/

Fixes: 07ce734dd8ad ("perf/x86/intel/uncore: Clean up client IMC")
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Link: https://lore.kernel.org/r/20220803160031.1379788-1-eranian@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 arch/x86/events/intel/uncore_snb.c |   18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)

--- a/arch/x86/events/intel/uncore_snb.c
+++ b/arch/x86/events/intel/uncore_snb.c
@@ -575,6 +575,22 @@ int snb_pci2phy_map_init(int devid)
 	return 0;
 }
 
+static u64 snb_uncore_imc_read_counter(struct intel_uncore_box *box, struct perf_event *event)
+{
+	struct hw_perf_event *hwc = &event->hw;
+
+	/*
+	 * SNB IMC counters are 32-bit and are laid out back to back
+	 * in MMIO space. Therefore we must use a 32-bit accessor function
+	 * using readq() from uncore_mmio_read_counter() causes problems
+	 * because it is reading 64-bit at a time. This is okay for the
+	 * uncore_perf_event_update() function because it drops the upper
+	 * 32-bits but not okay for plain uncore_read_counter() as invoked
+	 * in uncore_pmu_event_start().
+	 */
+	return (u64)readl(box->io_addr + hwc->event_base);
+}
+
 static struct pmu snb_uncore_imc_pmu = {
 	.task_ctx_nr	= perf_invalid_context,
 	.event_init	= snb_uncore_imc_event_init,
@@ -594,7 +610,7 @@ static struct intel_uncore_ops snb_uncor
 	.disable_event	= snb_uncore_imc_disable_event,
 	.enable_event	= snb_uncore_imc_enable_event,
 	.hw_config	= snb_uncore_imc_hw_config,
-	.read_counter	= uncore_mmio_read_counter,
+	.read_counter	= snb_uncore_imc_read_counter,
 };
 
 static struct intel_uncore_type snb_uncore_imc = {



  parent reply	other threads:[~2022-09-02 13:47 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-02 12:18 [PATCH 5.4 00/77] 5.4.212-rc1 review Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 01/77] audit: fix potential double free on error path from fsnotify_add_inode_mark Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 02/77] parisc: Fix exception handler for fldw and fstw instructions Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 03/77] kernel/sys_ni: add compat entry for fadvise64_64 Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 04/77] usb: cdns3: Fix issue for clear halt endpoint Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 05/77] Revert "selftests/bpf: Fix "dubious pointer arithmetic" test" Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 06/77] Revert "selftests/bpf: Fix test_align verifier log patterns" Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 07/77] pinctrl: amd: Dont save/restore interrupt status and wake status bits Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 08/77] sched/deadline: Unthrottle PI boosted threads while enqueuing Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 09/77] sched/deadline: Fix stale throttling on de-/boosted tasks Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 10/77] sched/deadline: Fix priority inheritance with multiple scheduling classes Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 11/77] kernel/sched: Remove dl_boosted flag comment Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 12/77] xfrm: fix refcount leak in __xfrm_policy_check() Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 13/77] af_key: Do not call xfrm_probe_algs in parallel Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 14/77] SUNRPC: RPC level errors should set task->tk_rpc_status Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 15/77] rose: check NULL rose_loopback_neigh->loopback Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 16/77] net/mlx5e: Properly disable vlan strip on non-UL reps Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 17/77] net: moxa: get rid of asymmetry in DMA mapping/unmapping Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 18/77] bonding: 802.3ad: fix no transmission of LACPDUs Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 19/77] net: ipvtap - add __init/__exit annotations to module init/exit funcs Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 20/77] netfilter: ebtables: reject blobs that dont provide all entry points Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 21/77] bnxt_en: fix NQ resource accounting during vf creation on 57500 chips Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 22/77] netfilter: nft_payload: report ERANGE for too long offset and length Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 23/77] netfilter: nft_payload: do not truncate csum_offset and csum_type Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 24/77] netfilter: nft_osf: restrict osf to ipv4, ipv6 and inet families Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 25/77] netfilter: nft_tunnel: restrict it to netdev family Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 26/77] net: Fix data-races around weight_p and dev_weight_[rt]x_bias Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 27/77] net: Fix data-races around netdev_tstamp_prequeue Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 28/77] ratelimit: Fix data-races in ___ratelimit() Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 29/77] net: Fix a data-race around sysctl_tstamp_allow_data Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 30/77] net: Fix a data-race around sysctl_net_busy_poll Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 31/77] net: Fix a data-race around sysctl_net_busy_read Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 32/77] net: Fix a data-race around netdev_budget Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 33/77] net: Fix a data-race around netdev_budget_usecs Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 34/77] net: Fix a data-race around sysctl_somaxconn Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 35/77] ixgbe: stop resetting SYSTIME in ixgbe_ptp_start_cyclecounter Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 36/77] btrfs: fix silent failure when deleting root reference Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 37/77] btrfs: replace: drop assert for suspended replace Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 38/77] btrfs: add info when mount fails due to stale replace target Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 39/77] btrfs: check if root is readonly while setting security xattr Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 40/77] x86/unwind/orc: Unwind ftrace trampolines with correct ORC entry Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 41/77] loop: Check for overflow while configuring loop Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 42/77] asm-generic: sections: refactor memory_intersects Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 43/77] s390: fix double free of GS and RI CBs on fork() failure Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 44/77] ACPI: processor: Remove freq Qos request for all CPUs Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 45/77] mm/hugetlb: fix hugetlb not supporting softdirty tracking Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 46/77] md: call __md_stop_writes in md_stop Greg Kroah-Hartman
2022-09-02 12:18 ` Greg Kroah-Hartman [this message]
2022-09-02 12:18 ` [PATCH 5.4 48/77] scsi: storvsc: Remove WQ_MEM_RECLAIM from storvsc_error_wq Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 49/77] mm: Force TLB flush for PFNMAP mappings before unlink_file_vma() Greg Kroah-Hartman
2022-09-02 12:18 ` [PATCH 5.4 50/77] s390/mm: do not trigger write fault when vma does not allow VM_WRITE Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 51/77] x86/bugs: Add "unknown" reporting for MMIO Stale Data Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 52/77] kbuild: Fix include path in scripts/Makefile.modpost Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 53/77] Bluetooth: L2CAP: Fix build errors in some archs Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 54/77] HID: steam: Prevent NULL pointer dereference in steam_{recv,send}_report Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 55/77] udmabuf: Set the DMA mask for the udmabuf device (v2) Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 56/77] media: pvrusb2: fix memory leak in pvr_probe Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 57/77] HID: hidraw: fix memory leak in hidraw_release() Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 58/77] fbdev: fb_pm2fb: Avoid potential divide by zero error Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 59/77] ftrace: Fix NULL pointer dereference in is_ftrace_trampoline when ftrace is dead Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 60/77] bpf: Dont redirect packets with invalid pkt_len Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 61/77] mm/rmap: Fix anon_vma->degree ambiguity leading to double-reuse Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 62/77] btrfs: introduce btrfs_lookup_match_dir Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 63/77] btrfs: do not pin logs too early during renames Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 64/77] btrfs: unify lookup return value when dir entry is missing Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 65/77] drm/amd/display: Avoid MPC infinite loop Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 66/77] drm/amd/display: clear optc underflow before turn off odm clock Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 67/77] neigh: fix possible DoS due to net iface start/stop loop Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 68/77] s390/hypfs: avoid error message under KVM Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 69/77] drm/amd/display: Fix pixel clock programming Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 70/77] netfilter: conntrack: NF_CONNTRACK_PROCFS should no longer default to y Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 71/77] btrfs: tree-checker: check for overlapping extent items Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 72/77] lib/vdso: Let do_coarse() return 0 to simplify the callsite Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 73/77] lib/vdso: Mark do_hres() and do_coarse() as __always_inline Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 74/77] kprobes: dont call disarm_kprobe() for disabled kprobes Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 75/77] io_uring: disable polling pollfree files Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 76/77] net/af_packet: check len when min_header_len equals to 0 Greg Kroah-Hartman
2022-09-02 12:19 ` [PATCH 5.4 77/77] net: neigh: dont call kfree_skb() under spin_lock_irqsave() Greg Kroah-Hartman
2022-09-02 17:07 ` [PATCH 5.4 00/77] 5.4.212-rc1 review Florian Fainelli
2022-09-02 22:16 ` Shuah Khan
2022-09-03  0:35 ` Guenter Roeck
2022-09-03 10:42 ` Sudip Mukherjee
2022-09-03 13:11 ` Naresh Kamboju

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220902121405.211947115@linuxfoundation.org \
    --to=gregkh@linuxfoundation.org \
    --cc=eranian@google.com \
    --cc=kan.liang@linux.intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=peterz@infradead.org \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox