- Kernel - mailweb.openeuler.org

[PATCH OLK-6.6] net/mlx5e: Move representor vnic reporter to eswitch devlink port
by Cai Xinchen 15 Oct '26

15 Oct '26

From: Carolina Jubran <cjubran(a)nvidia.com> mainline inclusion from mainline-v7.3-rc3 commit 7f26a5e8040b4957ef4dbdfcde6cc7ba2db53937 category: bugfix bugzilla: https://atomgit.com/src-openeuler/kernel/issues/20184 CVE: CVE-2026-97962 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id… -------------------------------- The representor vnic devlink health reporter is created and destroyed along the representor netdev (un)load path, which is not serialized by the devlink instance lock. Destroying the reporter from there triggers a devl_assert_locked() splat on driver unbind: WARNING: net/devlink/core.c:259 at devl_assert_locked+0x54/0x70, CPU#2: bash/3758 Modules linked in: mlx5_vdpa vringh vdpa mlx5_ib mlx5_fwctl mlx5_core ... CPU: 2 UID: 0 PID: 3758 Comm: bash Tainted: G W 6.19.0+ #1 PREEMPT Tainted: [W]=WARN Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), ... RIP: 0010:devl_assert_locked+0x54/0x70 Call Trace: <TASK> devl_health_reporter_destroy+0x3a/0x1b0 mlx5e_vport_rep_unload+0x12d/0x2b0 [mlx5_core] mlx5_eswitch_unregister_vport_reps+0x1b8/0x220 [mlx5_core] ? __esw_offloads_unload_rep+0x190/0x190 [mlx5_core] ? kernfs_remove_by_name_ns+0xc3/0xf0 device_release_driver_internal+0x3b2/0x560 unbind_store+0xce/0xf0 Move the reporter's lifecycle to the eswitch devlink port (un)register paths, which are already serialized by the devlink instance lock, and store the handle on mlx5_devlink_port. Use the port's mlx5_vport as the reporter priv since the diagnose callback only needs a device handle and a vport number, and mlx5_vport carries both and is initialized before any representor driver probes. Fixes: cf14af140a5a ("net/mlx5e: Add vnic devlink health reporter to representors") Signed-off-by: Carolina Jubran <cjubran(a)nvidia.com> Reviewed-by: Cosmin Ratiu <cratiu(a)nvidia.com> Signed-off-by: Tariq Toukan <tariqt(a)nvidia.com> Reviewed-by: Simon Horman <horms(a)kernel.org> Link: https://patch.msgid.link/20260906090700.3761260-1-tariqt@nvidia.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> Conflicts: drivers/net/ethernet/mellanox/mlx5/core/en_rep.c drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c [Commit 7f26a5e8040b4 built on newer mainline context not present in this tree: devl_port_health_reporter_create() still takes a graceful_period argument in 6.6, so pass 0 like the old en_rep.c code did, and the mlx5_esw_devlink_port_res_register/unregister() calls in the upstream context do not exist here, so drop them; the local "devlink.h" include is not needed either.] Signed-off-by: Cai Xinchen <caixinchen1(a)huawei.com> --- .../net/ethernet/mellanox/mlx5/core/en_rep.c | 51 +------------------ .../net/ethernet/mellanox/mlx5/core/en_rep.h | 1 - .../mellanox/mlx5/core/esw/devlink_port.c | 36 +++++++++++++ .../net/ethernet/mellanox/mlx5/core/eswitch.h | 1 + 4 files changed, 38 insertions(+), 51 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c index ce2151d98af9..71b7aff78ccd 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c @@ -54,7 +54,6 @@ #include "lib/vxlan.h" #define CREATE_TRACE_POINTS #include "diag/en_rep_tracepoint.h" -#include "diag/reporter_vnic.h" #include "en_accel/ipsec.h" #include "en/tc/int_port.h" #include "en/ptp.h" @@ -1364,50 +1363,6 @@ static unsigned int mlx5e_ul_rep_stats_grps_num(struct mlx5e_priv *priv) return ARRAY_SIZE(mlx5e_ul_rep_stats_grps); } -static int -mlx5e_rep_vnic_reporter_diagnose(struct devlink_health_reporter *reporter, - struct devlink_fmsg *fmsg, - struct netlink_ext_ack *extack) -{ - struct mlx5e_rep_priv *rpriv = devlink_health_reporter_priv(reporter); - struct mlx5_eswitch_rep *rep = rpriv->rep; - - return mlx5_reporter_vnic_diagnose_counters(rep->esw->dev, fmsg, - rep->vport, true); -} - -static const struct devlink_health_reporter_ops mlx5_rep_vnic_reporter_ops = { - .name = "vnic", - .diagnose = mlx5e_rep_vnic_reporter_diagnose, -}; - -static void mlx5e_rep_vnic_reporter_create(struct mlx5e_priv *priv, - struct devlink_port *dl_port) -{ - struct mlx5e_rep_priv *rpriv = priv->ppriv; - struct devlink_health_reporter *reporter; - - reporter = devl_port_health_reporter_create(dl_port, - &mlx5_rep_vnic_reporter_ops, - 0, rpriv); - if (IS_ERR(reporter)) { - mlx5_core_err(priv->mdev, - "Failed to create representor vnic reporter, err = %ld\n", - PTR_ERR(reporter)); - return; - } - - rpriv->rep_vnic_reporter = reporter; -} - -static void mlx5e_rep_vnic_reporter_destroy(struct mlx5e_priv *priv) -{ - struct mlx5e_rep_priv *rpriv = priv->ppriv; - - if (!IS_ERR_OR_NULL(rpriv->rep_vnic_reporter)) - devl_health_reporter_destroy(rpriv->rep_vnic_reporter); -} - static const struct mlx5e_profile mlx5e_rep_profile = { .init = mlx5e_init_rep, .cleanup = mlx5e_cleanup_rep, @@ -1516,10 +1471,8 @@ mlx5e_vport_vf_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep) dl_port = mlx5_esw_offloads_devlink_port(dev->priv.eswitch, rpriv->rep->vport); - if (!IS_ERR(dl_port)) { + if (!IS_ERR(dl_port)) SET_NETDEV_DEVLINK_PORT(netdev, dl_port); - mlx5e_rep_vnic_reporter_create(priv, dl_port); - } err = register_netdev(netdev); if (err) { @@ -1532,7 +1485,6 @@ mlx5e_vport_vf_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep) return 0; err_detach_netdev: - mlx5e_rep_vnic_reporter_destroy(priv); mlx5e_detach_netdev(netdev_priv(netdev)); err_cleanup_profile: priv->profile->cleanup(priv); @@ -1590,7 +1542,6 @@ mlx5e_vport_rep_unload(struct mlx5_eswitch_rep *rep) } unregister_netdev(netdev); - mlx5e_rep_vnic_reporter_destroy(priv); mlx5e_detach_netdev(priv); priv->profile->cleanup(priv); mlx5e_destroy_netdev(netdev); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h index 70640fa1ad7b..bcd7b4e814d0 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h @@ -118,7 +118,6 @@ struct mlx5e_rep_priv { struct rtnl_link_stats64 prev_vf_vport_stats; struct mlx5_flow_handle *send_to_vport_meta_rule; struct rhashtable tc_ht; - struct devlink_health_reporter *rep_vnic_reporter; }; static inline diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c index 91319b5acd3d..4c4cdd62f5dc 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c @@ -3,6 +3,26 @@ #include <linux/mlx5/driver.h> #include "eswitch.h" +#include "diag/reporter_vnic.h" + +static int +mlx5_esw_rep_vnic_reporter_diagnose(struct devlink_health_reporter *reporter, + struct devlink_fmsg *fmsg, + struct netlink_ext_ack *extack) +{ + struct mlx5_vport *vport = devlink_health_reporter_priv(reporter); + + mlx5_reporter_vnic_diagnose_counters(vport->dev, fmsg, vport->vport, + true); + + return 0; +} + +static const +struct devlink_health_reporter_ops mlx5_esw_rep_vnic_reporter_ops = { + .name = "vnic", + .diagnose = mlx5_esw_rep_vnic_reporter_diagnose, +}; static void mlx5_esw_get_port_parent_id(struct mlx5_core_dev *dev, struct netdev_phys_item_id *ppid) @@ -149,6 +169,7 @@ static const struct devlink_port_ops mlx5_esw_dl_sf_port_ops = { int mlx5_esw_offloads_devlink_port_register(struct mlx5_eswitch *esw, struct mlx5_vport *vport) { + struct devlink_health_reporter *reporter; struct mlx5_core_dev *dev = esw->dev; const struct devlink_port_ops *ops; struct mlx5_devlink_port *dl_port; @@ -178,6 +199,16 @@ int mlx5_esw_offloads_devlink_port_register(struct mlx5_eswitch *esw, struct mlx if (err) goto rate_err; + reporter = devl_port_health_reporter_create( + &dl_port->dl_port, &mlx5_esw_rep_vnic_reporter_ops, + 0, vport); + if (IS_ERR(reporter)) + mlx5_core_err(dev, + "Failed to create vnic health reporter for vport %d: %pe\n", + vport_num, reporter); + else + dl_port->vnic_reporter = reporter; + return 0; rate_err: @@ -193,6 +224,11 @@ void mlx5_esw_offloads_devlink_port_unregister(struct mlx5_eswitch *esw, struct return; dl_port = vport->dl_port; + if (dl_port->vnic_reporter) { + devl_health_reporter_destroy(dl_port->vnic_reporter); + dl_port->vnic_reporter = NULL; + } + mlx5_esw_qos_vport_update_group(esw, vport, NULL, NULL); devl_rate_leaf_destroy(&dl_port->dl_port); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h index d5ebd4f01103..a47692141f8f 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h @@ -179,6 +179,7 @@ struct mlx5_vport; struct mlx5_devlink_port { struct devlink_port dl_port; struct mlx5_vport *vport; + struct devlink_health_reporter *vnic_reporter; }; static inline void mlx5_devlink_port_init(struct mlx5_devlink_port *dl_port, -- 2.18.0.huawei.25

2 1

[PATCH OLK-6.6] nfsd: split nfsd4_copy into transient and durable async copy objects
by Lu Jialin 15 Oct '26

15 Oct '26

From: Jeff Layton <jlayton(a)kernel.org> mainline inclusion from mainline-v7.3-rc1 commit f307b4f7ed174a3252d2f292904fd276c78c77de category: bugfix bugzilla: https://atomgit.com/src-openeuler/kernel/issues/18939 CVE: CVE-2026-89676 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- struct nfsd4_copy served two roles: as &u->copy it is a transient per-COMPOUND argument in the request buffer; as the heap async_copy it is a durable object (worker kthread, reaper linkage, CB_OFFLOAD callback, IDR stateid) that outlives the COMPOUND, with dup_copy_fields() shuttling state between them. That dual identity was the root of the recent lifetime bugs. Introduce struct nfsd4_async_copy for the durable object. It embeds a struct nfsd4_copy (cp_copy) for the operation parameters/result and adds the durable-only fields: async_copies linkage, task_struct, refcount, reaper TTL, copy stateid, and CB_OFFLOAD callback. The durable object therefore never points into the request buffer. cp_clp stays in nfsd4_copy -- it is a request property read by the sync-copy tracepoints on the transient object. Mechanical split, no intended behavioral change; a step toward folding the copy stateids into the common nfs4_stid infrastructure. Assisted-by: Claude:claude-opus-4-8 Signed-off-by: Jeff Layton <jlayton(a)kernel.org> Link: https://patch.msgid.link/20260710-nfsd-testing-v3-8-a0ff7db6aa3e@kernel.org Signed-off-by: Chuck Lever <cel(a)kernel.org> Conflicts: fs/nfsd/nfs4proc.c [OLK-6.6 is a v6.6-era baseline that lacks the nfsd async-copy prerequisites present in the mainline base (v7.3-rc1): there is no nfsd4_has_active_async_copies(), nfsd4_async_copy_reaper() or nfsd4_cancel_copy_by_sb() to repoint and no cp_ttl reaper, so those hunks do not apply; nfs4_put_copy()/nfsd4_stop_copy()/ nfsd4_unhash_copy() keep OLK-6.6's earlier forms (no task_struct pinning in nfs4_put_copy(), plain cp_clp store, no CB_ERROR/COMPLETED handling) with only the type and cp_copy field paths adjusted; cleanup_async_copy() additionally keeps OLK-6.6's 'if (cp_copy.cp_clp) list_del' block] fs/nfsd/nfs4state.c [nfs4_init_copy_state()/nfs4_free_copy_state() repointed to struct nfsd4_async_copy; nfs4_init_cp_state() keeps the 3-argument form used by OLK-6.6 (mainline's base added a struct nfs4_stid *p_stid param)] fs/nfsd/state.h [forward declaration added for struct nfsd4_async_copy and the two nfs4_*_copy_state() prototypes repointed to it] fs/nfsd/xdr4.h [OLK-6.6's struct nfsd4_copy has no cp_cb_offload, cp_ttl or attr_update (mainline added them in earlier patches), so only copies/copy_task/refcount/cp_stateid move into the new struct nfsd4_async_copy; OLK-6.6 still heap-allocates the CB_OFFLOAD with kzalloc() and uses nfsd4_send_cb_offload(copy, nfserr), so cp_cb_offload is not embedded and nfsd4_cb_offload_release() needs no change] Signed-off-by: Lu Jialin <lujialin4(a)huawei.com> --- fs/nfsd/nfs4proc.c | 89 ++++++++++++++++++++++++--------------------- fs/nfsd/nfs4state.c | 6 +-- fs/nfsd/state.h | 5 ++- fs/nfsd/xdr4.h | 23 +++++++++--- 4 files changed, 70 insertions(+), 53 deletions(-) diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c index c74aedc1d2c6..ae3ad8afaffd 100644 --- a/fs/nfsd/nfs4proc.c +++ b/fs/nfsd/nfs4proc.c @@ -1274,32 +1274,32 @@ nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, return status; } -static void nfs4_put_copy(struct nfsd4_copy *copy) +static void nfs4_put_copy(struct nfsd4_async_copy *copy) { if (!refcount_dec_and_test(&copy->refcount)) return; - atomic_dec(&copy->cp_nn->pending_async_copies); - kfree(copy->cp_src); + atomic_dec(&copy->cp_copy.cp_nn->pending_async_copies); + kfree(copy->cp_copy.cp_src); kfree(copy); } -static void nfsd4_stop_copy(struct nfsd4_copy *copy) +static void nfsd4_stop_copy(struct nfsd4_async_copy *copy) { - if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags)) + if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_copy.cp_flags)) kthread_stop(copy->copy_task); nfs4_put_copy(copy); } -static struct nfsd4_copy *nfsd4_unhash_copy(struct nfs4_client *clp) +static struct nfsd4_async_copy *nfsd4_unhash_copy(struct nfs4_client *clp) { - struct nfsd4_copy *copy = NULL; + struct nfsd4_async_copy *copy = NULL; spin_lock(&clp->async_lock); if (!list_empty(&clp->async_copies)) { - copy = list_first_entry(&clp->async_copies, struct nfsd4_copy, - copies); + copy = list_first_entry(&clp->async_copies, + struct nfsd4_async_copy, copies); refcount_inc(&copy->refcount); - copy->cp_clp = NULL; + copy->cp_copy.cp_clp = NULL; if (!list_empty(&copy->copies)) list_del_init(&copy->copies); } @@ -1309,7 +1309,7 @@ static struct nfsd4_copy *nfsd4_unhash_copy(struct nfs4_client *clp) void nfsd4_shutdown_copy(struct nfs4_client *clp) { - struct nfsd4_copy *copy; + struct nfsd4_async_copy *copy; while ((copy = nfsd4_unhash_copy(clp)) != NULL) nfsd4_stop_copy(copy); @@ -1702,7 +1702,6 @@ static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst) if (!nfsd4_ssc_is_inter(src)) dst->nf_src = nfsd_file_get(src->nf_src); - memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid)); memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server)); memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid)); memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh)); @@ -1717,20 +1716,20 @@ static void release_copy_files(struct nfsd4_copy *copy) nfsd_file_put(copy->nf_dst); } -static void cleanup_async_copy(struct nfsd4_copy *copy) +static void cleanup_async_copy(struct nfsd4_async_copy *copy) { nfs4_free_copy_state(copy); - release_copy_files(copy); - if (copy->cp_clp) { - spin_lock(&copy->cp_clp->async_lock); + release_copy_files(&copy->cp_copy); + if (copy->cp_copy.cp_clp) { + spin_lock(&copy->cp_copy.cp_clp->async_lock); if (!list_empty(&copy->copies)) list_del_init(&copy->copies); - spin_unlock(&copy->cp_clp->async_lock); + spin_unlock(&copy->cp_copy.cp_clp->async_lock); } nfs4_put_copy(copy); } -static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr) +static void nfsd4_send_cb_offload(struct nfsd4_async_copy *copy, __be32 nfserr) { struct nfsd4_cb_offload *cbo; @@ -1738,14 +1737,14 @@ static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr) if (!cbo) return; - memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res)); - memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh)); + memcpy(&cbo->co_res, &copy->cp_copy.cp_res, sizeof(copy->cp_copy.cp_res)); + memcpy(&cbo->co_fh, &copy->cp_copy.fh, sizeof(copy->cp_copy.fh)); cbo->co_nfserr = nfserr; - nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops, + nfsd4_init_cb(&cbo->co_cb, copy->cp_copy.cp_clp, &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD); - trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid, - &cbo->co_fh, copy->cp_count, nfserr); + trace_nfsd_cb_offload(copy->cp_copy.cp_clp, &cbo->co_res.cb_stateid, + &cbo->co_fh, copy->cp_copy.cp_count, nfserr); nfsd4_run_cb(&cbo->co_cb); } @@ -1758,7 +1757,8 @@ static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr) */ static int nfsd4_do_async_copy(void *data) { - struct nfsd4_copy *copy = (struct nfsd4_copy *)data; + struct nfsd4_async_copy *async = data; + struct nfsd4_copy *copy = &async->cp_copy; __be32 nfserr; if (nfsd4_ssc_is_inter(copy)) { @@ -1786,8 +1786,8 @@ static int nfsd4_do_async_copy(void *data) } do_callback: - nfsd4_send_cb_offload(copy, nfserr); - cleanup_async_copy(copy); + nfsd4_send_cb_offload(async, nfserr); + cleanup_async_copy(async); return 0; } @@ -1796,7 +1796,7 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, union nfsd4_op_u *u) { struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id); - struct nfsd4_copy *async_copy = NULL; + struct nfsd4_async_copy *async_copy = NULL; struct nfsd4_copy *copy = &u->copy; struct nfsd42_write_res *result; __be32 status; @@ -1822,33 +1822,38 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, memcpy(&copy->fh, &cstate->current_fh.fh_handle, sizeof(struct knfsd_fh)); if (nfsd4_copy_is_async(copy)) { - async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL); + async_copy = kzalloc(sizeof(struct nfsd4_async_copy), GFP_KERNEL); if (!async_copy) goto out_err; - async_copy->cp_nn = nn; + async_copy->cp_copy.cp_nn = nn; INIT_LIST_HEAD(&async_copy->copies); refcount_set(&async_copy->refcount, 1); /* Arbitrary cap on number of pending async copy operations */ if (atomic_inc_return(&nn->pending_async_copies) > (int)rqstp->rq_pool->sp_nrthreads) goto out_err; - async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL); - if (!async_copy->cp_src) + async_copy->cp_copy.cp_src = + kmalloc(sizeof(*async_copy->cp_copy.cp_src), GFP_KERNEL); + if (!async_copy->cp_copy.cp_src) goto out_err; - dup_copy_fields(copy, async_copy); if (!nfs4_init_copy_state(nn, async_copy)) goto out_err; memcpy(&result->cb_stateid, &async_copy->cp_stateid.cs_stid, sizeof(result->cb_stateid)); + /* + * dup after writing cb_stateid; duplicating first would leave + * the callback stateid zeroed. + */ + dup_copy_fields(copy, &async_copy->cp_copy); async_copy->copy_task = kthread_create(nfsd4_do_async_copy, async_copy, "%s", "copy thread"); if (IS_ERR(async_copy->copy_task)) goto out_err; - spin_lock(&async_copy->cp_clp->async_lock); + spin_lock(&async_copy->cp_copy.cp_clp->async_lock); list_add(&async_copy->copies, - &async_copy->cp_clp->async_copies); - spin_unlock(&async_copy->cp_clp->async_lock); + &async_copy->cp_copy.cp_clp->async_copies); + spin_unlock(&async_copy->cp_copy.cp_clp->async_lock); wake_up_process(async_copy->copy_task); status = nfs_ok; } else { @@ -1873,10 +1878,10 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, goto out; } -static struct nfsd4_copy * +static struct nfsd4_async_copy * find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid) { - struct nfsd4_copy *copy; + struct nfsd4_async_copy *copy; lockdep_assert_held(&clp->async_lock); @@ -1888,10 +1893,10 @@ find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid) return NULL; } -static struct nfsd4_copy * +static struct nfsd4_async_copy * find_async_copy(struct nfs4_client *clp, stateid_t *stateid) { - struct nfsd4_copy *copy; + struct nfsd4_async_copy *copy; spin_lock(&clp->async_lock); copy = find_async_copy_locked(clp, stateid); @@ -1907,7 +1912,7 @@ nfsd4_offload_cancel(struct svc_rqst *rqstp, union nfsd4_op_u *u) { struct nfsd4_offload_status *os = &u->offload_status; - struct nfsd4_copy *copy; + struct nfsd4_async_copy *copy; struct nfs4_client *clp = cstate->clp; copy = find_async_copy(clp, &os->stateid); @@ -1995,13 +2000,13 @@ nfsd4_offload_status(struct svc_rqst *rqstp, { struct nfsd4_offload_status *os = &u->offload_status; __be32 status = nfs_ok; - struct nfsd4_copy *copy; + struct nfsd4_async_copy *copy; struct nfs4_client *clp = cstate->clp; spin_lock(&clp->async_lock); copy = find_async_copy_locked(clp, &os->stateid); if (copy) - os->count = copy->cp_res.wr_bytes_written; + os->count = copy->cp_copy.cp_res.wr_bytes_written; else status = nfserr_bad_stateid; spin_unlock(&clp->async_lock); diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index 5140197771ad..0d62c65ae4d4 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -1034,7 +1034,7 @@ static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid, return 1; } -int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy) +int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_async_copy *copy) { return nfs4_init_cp_state(nn, &copy->cp_stateid, NFS4_COPY_STID); } @@ -1068,13 +1068,13 @@ struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn, return NULL; } -void nfs4_free_copy_state(struct nfsd4_copy *copy) +void nfs4_free_copy_state(struct nfsd4_async_copy *copy) { struct nfsd_net *nn; if (copy->cp_stateid.cs_type != NFS4_COPY_STID) return; - nn = net_generic(copy->cp_clp->net, nfsd_net_id); + nn = net_generic(copy->cp_copy.cp_clp->net, nfsd_net_id); spin_lock(&nn->s2s_cp_lock); idr_remove(&nn->s2s_cp_stateids, copy->cp_stateid.cs_stid.si_opaque.so_id); diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h index 4dd8ade17446..d62faf4f183a 100644 --- a/fs/nfsd/state.h +++ b/fs/nfsd/state.h @@ -676,6 +676,7 @@ struct nfsd4_blocked_lock { struct nfsd4_compound_state; struct nfsd_net; struct nfsd4_copy; +struct nfsd4_async_copy; extern __be32 nfs4_preprocess_stateid_op(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct svc_fh *fhp, @@ -686,8 +687,8 @@ __be32 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate, struct nfs4_stid **s, struct nfsd_net *nn); struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab, void (*sc_free)(struct nfs4_stid *)); -int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy); -void nfs4_free_copy_state(struct nfsd4_copy *copy); +int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_async_copy *copy); +void nfs4_free_copy_state(struct nfsd4_async_copy *copy); struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn, struct nfs4_stid *p_stid); void nfs4_unhash_stid(struct nfs4_stid *s); diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h index a83a52b0f518..4ac30b2ec7c5 100644 --- a/fs/nfsd/xdr4.h +++ b/fs/nfsd/xdr4.h @@ -599,18 +599,29 @@ struct nfsd4_copy { struct nfsd_file *nf_src; struct nfsd_file *nf_dst; - copy_stateid_t cp_stateid; - - struct list_head copies; - struct task_struct *copy_task; - refcount_t refcount; - struct nfsd4_ssc_umount_item *ss_nsui; struct nfs_fh c_fh; nfs4_stateid stateid; struct nfsd_net *cp_nn; }; +/* + * Durable state for an async (background) server-side COPY. + * + * struct nfsd4_copy is transient: it lives in the COMPOUND argument buffer + * and is reused once the op returns. An async COPY outlives the COMPOUND + * (worker kthread, reaper linkage), so its params and result are snapshotted + * into the embedded cp_copy and it never points into the request buffer. + */ +struct nfsd4_async_copy { + struct nfsd4_copy cp_copy; /* operation params + result */ + + struct list_head copies; /* nfs4_client.async_copies */ + struct task_struct *copy_task; + refcount_t refcount; + copy_stateid_t cp_stateid; /* s2s_cp_stateids IDR entry */ +}; + static inline void nfsd4_copy_set_sync(struct nfsd4_copy *copy, bool sync) { if (sync) -- 2.34.1

2 1

[PATCH OLK-6.6] kabi: Add a Python3 port of check-kabi
by GONG Ruiqi 15 Oct '26

15 Oct '26

hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/10129 ---------------------------------------- scripts/check-kabi is written in Python 2, which has long been in End-Of-Life[1], and its runtime environment is becoming increasingly difficult to obtain. Port it as scripts/check-kabi-py3 with the behavior kept identical: the same -k/-s options, the same output text and the same exit codes (0 on a clean check, 1 on kABI breakage, moved symbols or usage errors). The two byte-identical loaders of the original are folded into a single load_table() helper and the unused os/re imports are dropped along the way; neither change is observable from outside. Link: https://devguide.python.org/versions/ [1] Assisted-by: ZCode:GLM-5.3 Signed-off-by: GONG Ruiqi <gongruiqi1(a)huawei.com> --- scripts/check-kabi-py3 | 114 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 114 insertions(+) create mode 100755 scripts/check-kabi-py3 diff --git a/scripts/check-kabi-py3 b/scripts/check-kabi-py3 new file mode 100755 index 000000000000..3dac8fdda950 --- /dev/null +++ b/scripts/check-kabi-py3 @@ -0,0 +1,114 @@ +#!/usr/bin/env python3 +# SPDX-License-Identifier: GPL-2.0 +# +# check-kabi - Red Hat kABI reference checking tool +# +# We use this script to check against reference Module.kabi files. +# +# Author: Jon Masters <jcm(a)redhat.com> +# Copyright (C) 2007-2009 Red Hat, Inc. +# +# Python 2 is no longer maintained, hence the port to Python 3. +# +# Author: GONG Ruiqi <gongruiqi1(a)huawei.com> +# Copyright (C) 2026 Huawei Technologies Co., Ltd. +# +# This software may be freely redistributed under the terms of the GNU +# General Public License (GPL). + +import getopt +import sys + +def load_table(table, filename): + """Load a Module.symvers or Module.kabi file.""" + + with open(filename, "r") as f: + for in_line in f: + if in_line == "\n": + continue + checksum, symbol, directory, type, namespace = in_line.split("\t") + table[symbol] = in_line[0:-1] + +def check_kabi(symvers, kabi): + """Check Module.kabi and Module.symvers files.""" + + fail = False + warn = False + changed_symbols = [] + moved_symbols = [] + + for symbol, line in kabi.items(): + abi_hash, _, abi_dir, _, _ = line.split("\t") + if symbol in symvers: + sym_hash, _, sym_dir, _, _ = symvers[symbol].split("\t") + if abi_hash != sym_hash: + fail = True + changed_symbols.append(symbol) + + if abi_dir != sym_dir: + warn = True + moved_symbols.append(symbol) + else: + fail = True + changed_symbols.append(symbol) + + if fail: + print("*** ERROR - ABI BREAKAGE WAS DETECTED ***") + print("") + print("The following symbols have been changed (this will cause an ABI breakage):") + print("") + for symbol in changed_symbols: + print(symbol) + print("") + + if warn: + print("*** WARNING - ABI SYMBOLS MOVED ***") + print("") + print("The following symbols moved (typically caused by moving a symbol from being") + print("provided by the kernel vmlinux out to a loadable module):") + print("") + for symbol in moved_symbols: + print(symbol) + print("") + + # Halt the build, if we got errors and/or warnings. In either case, + # double-checking is required to avoid introducing / concealing + # KABI inconsistencies. + if fail or warn: + sys.exit(1) + sys.exit(0) + +def usage(): + print(""" +check-kabi-py3: check Module.kabi and Module.symvers files. + + check-kabi-py3 [ -k Module.kabi ] [ -s Module.symvers ] + +""") + +if __name__ == "__main__": + + symvers_file = "" + kabi_file = "" + + opts, args = getopt.getopt(sys.argv[1:], "hk:s:") + + for o, v in opts: + if o == "-s": + symvers_file = v + if o == "-h": + usage() + sys.exit(0) + if o == "-k": + kabi_file = v + + if (symvers_file == "") or (kabi_file == ""): + usage() + sys.exit(1) + + symvers = {} + kabi = {} + + load_table(symvers, symvers_file) + load_table(kabi, kabi_file) + check_kabi(symvers, kabi) -- 2.43.0

2 1

[PATCH OLK-6.6] ALSA: pcm: set timer->private_data before registering the PCM timer
by Pan Taixi 15 Oct '26

15 Oct '26

From: Nguyen Ngoc Thang <ngocthang2710.1999(a)gmail.com> stable inclusion from stable-v6.6.158 commit 686c7a6af1eea8d2a919303e4c6bbec3de79dbdc category: bugfix bugzilla: https://atomgit.com/src-openeuler/kernel/issues/20047 CVE: CVE-2026-98314 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id… -------------------------------- ALSA: pcm: set timer->private_data before registering the PCM timer [ Upstream commit 1e713f9bb2ac583521f06b0eb4e22440b1e3d078 ] snd_pcm_timer_init() calls snd_device_register() to link the new struct snd_timer into the global timer list while it still carries hw.c_resolution = snd_pcm_timer_resolution (and hw.start/hw.stop), and only afterwards sets timer->private_data = substream. Once the timer is on the list under register_mutex, a concurrent reader can already reach it through the same mutex and invoke these callbacks. /proc/asound/timers does this via c_resolution(), and snd_timer_open()+snd_timer_start() reach start()/stop() the same way. All three dereference timer->private_data, which for this brief window is NULL, giving a NULL-pointer dereference: substream = timer->private_data; return substream->runtime ? ... // substream is NULL Move the private_data/private_free assignment before snd_device_register() so the timer is never visible on the list without its private_data set. On the snd_device_register() failure path, private_free() (snd_pcm_timer_free()) can now run, but it only does substream->timer = NULL, which is already NULL at that point since substream->timer is set to the new timer just once, after a successful registration -- so the failure path stays safe. Reported-by: syzbot+19da64013c46df87f971(a)syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=19da64013c46df87f971 Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Nguyen Ngoc Thang <ngocthang2710.1999(a)gmail.com> Link: https://patch.msgid.link/20260913134446.114724-1-ngocthang2710.1999@gmail.c… Signed-off-by: Takashi Iwai <tiwai(a)suse.de> Signed-off-by: Sasha Levin <sashal(a)kernel.org> Signed-off-by: Pan Taixi <pantaixi1(a)huawei.com> --- sound/core/pcm_timer.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/sound/core/pcm_timer.c b/sound/core/pcm_timer.c index c43484b22b34..725a7272cd07 100644 --- a/sound/core/pcm_timer.c +++ b/sound/core/pcm_timer.c @@ -110,16 +110,19 @@ void snd_pcm_timer_init(struct snd_pcm_substream *substream) sprintf(timer->name, "PCM %s %i-%i-%i", substream->stream == SNDRV_PCM_STREAM_CAPTURE ? "capture" : "playback", tid.card, tid.device, tid.subdevice); timer->hw = snd_pcm_timer; + /* Set before registering: a concurrent reader can invoke our hw + * callbacks as soon as the timer is on the global list. + */ + timer->private_data = substream; + timer->private_free = snd_pcm_timer_free; if (snd_device_register(timer->card, timer) < 0) { snd_device_free(timer->card, timer); return; } - timer->private_data = substream; - timer->private_free = snd_pcm_timer_free; substream->timer = timer; } void snd_pcm_timer_done(struct snd_pcm_substream *substream) { -- 2.34.1

2 1

[PATCH OLK-5.10] ALSA: pcm: set timer->private_data before registering the PCM timer
by Pan Taixi 15 Oct '26

15 Oct '26

From: Nguyen Ngoc Thang <ngocthang2710.1999(a)gmail.com> stable inclusion from stable-v5.10.271 commit 27f167e117eca310661fbcbacb352ea08face067 category: bugfix bugzilla: https://atomgit.com/src-openeuler/kernel/issues/20047 CVE: CVE-2026-98314 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id… -------------------------------- ALSA: pcm: set timer->private_data before registering the PCM timer [ Upstream commit 1e713f9bb2ac583521f06b0eb4e22440b1e3d078 ] snd_pcm_timer_init() calls snd_device_register() to link the new struct snd_timer into the global timer list while it still carries hw.c_resolution = snd_pcm_timer_resolution (and hw.start/hw.stop), and only afterwards sets timer->private_data = substream. Once the timer is on the list under register_mutex, a concurrent reader can already reach it through the same mutex and invoke these callbacks. /proc/asound/timers does this via c_resolution(), and snd_timer_open()+snd_timer_start() reach start()/stop() the same way. All three dereference timer->private_data, which for this brief window is NULL, giving a NULL-pointer dereference: substream = timer->private_data; return substream->runtime ? ... // substream is NULL Move the private_data/private_free assignment before snd_device_register() so the timer is never visible on the list without its private_data set. On the snd_device_register() failure path, private_free() (snd_pcm_timer_free()) can now run, but it only does substream->timer = NULL, which is already NULL at that point since substream->timer is set to the new timer just once, after a successful registration -- so the failure path stays safe. Reported-by: syzbot+19da64013c46df87f971(a)syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=19da64013c46df87f971 Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Nguyen Ngoc Thang <ngocthang2710.1999(a)gmail.com> Link: https://patch.msgid.link/20260913134446.114724-1-ngocthang2710.1999@gmail.c… Signed-off-by: Takashi Iwai <tiwai(a)suse.de> Signed-off-by: Sasha Levin <sashal(a)kernel.org> Signed-off-by: Pan Taixi <pantaixi1(a)huawei.com> --- sound/core/pcm_timer.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/sound/core/pcm_timer.c b/sound/core/pcm_timer.c index c43484b22b34..725a7272cd07 100644 --- a/sound/core/pcm_timer.c +++ b/sound/core/pcm_timer.c @@ -110,16 +110,19 @@ void snd_pcm_timer_init(struct snd_pcm_substream *substream) sprintf(timer->name, "PCM %s %i-%i-%i", substream->stream == SNDRV_PCM_STREAM_CAPTURE ? "capture" : "playback", tid.card, tid.device, tid.subdevice); timer->hw = snd_pcm_timer; + /* Set before registering: a concurrent reader can invoke our hw + * callbacks as soon as the timer is on the global list. + */ + timer->private_data = substream; + timer->private_free = snd_pcm_timer_free; if (snd_device_register(timer->card, timer) < 0) { snd_device_free(timer->card, timer); return; } - timer->private_data = substream; - timer->private_free = snd_pcm_timer_free; substream->timer = timer; } void snd_pcm_timer_done(struct snd_pcm_substream *substream) { -- 2.34.1

2 1

openEuler Kernel SIG双周例会
by openEuler conference 14 Oct '26

14 Oct '26

您好！ Kernel 邀请您参加 2026-10-16 14:00 召开的WeLink会议(自动录制) 会议主题：openEuler Kernel SIG双周例会会议发起人： LiaoTao_Wave 会议内容： 1. 进展update 2. 议题征集中（新增议题可直接填报至会议纪要看板）会议链接：https://meeting.huaweicloud.com:36443/#/j/986389680 会议纪要&签到链接：https://etherpad.openeuler.org/p/Kernel-meetings 更多资讯尽在：https://www.openeuler.org/zh/ Hello! Kernel invites you to attend the WeLink conference(auto recording) will be held at 2026-10-16 14:00 The subject of the conference is openEuler Kernel SIG双周例会 The sponsor of the conference is LiaoTao_Wave Summary: 1. 进展update 2. 议题征集中（新增议题可直接填报至会议纪要看板） You can join the meeting at https://meeting.huaweicloud.com:36443/#/j/986389680 Add topics at https://etherpad.openeuler.org/p/Kernel-meetings More information: https://www.openeuler.org/en/

1 0

[PATCH OLK-6.6 0/8] Add support for vTLBID
by Jinqian Yang 14 Oct '26

14 Oct '26

This series adds KVM/arm64 support for the ARM TLB invalidation domain (TLBID) extension. With TLBID a guest can scope its TLB invalidation broadcasts to a virtual domain (vDomain). User space discovers the feature through the new KVM_CAP_ARM_TLBIDOMAIN capability. The number of supported vDomains is queried with the new VM ioctl KVM_ARM_GET_VDOMAIN_NUM, and the per-vCPU ioctl KVM_ARM_VCPU_SET_VDOMAIN tells KVM which vDomains each vCPU belongs to, via a vDomain bitmap. KVM maintains a per-VM mapping from vDomains to the physical CPUs that run their member vCPUs, and reprogrammes the VTLBID(n)_EL2 registers on every vCPU load, so that a guest TLBI tagged with a vDomain ID only reaches the physical CPUs currently running that vDomain's vCPUs. Because DVMBM and TLBID are mutually exclusive, DVMBM is force-disabled once TLBID is enabled for a VM. On the guest side, TLBIDIDR_EL1 is emulated with virtual domain counts clamped to the hardware limits, the ID_AA64MMFR4_EL1.TLBID field is exposed only when the feature is active for the VM, and guest accesses to TLBIDIDR_EL1 are trapped and handled by KVM. The feature is gated behind CONFIG_KVM_ARM_VTLBID \and additionally requires hardware TLBID support at run time. Tracepoints are added for the vCPU load path so that the vDomain-to-pCPU mapping can be observed. Jinqian Yang (8): KVM: arm64: Introduce TLBID virtualization infrastructure KVM: arm64: Add VTLBID sysregs and update VTLBID(n) on vCPU load KVM: arm64: Wire up TLBID vDomain configuration ioctls KVM: arm64: Ensure DVMBM is disabled when TLBID is enabled KVM: arm64: Emulate TLBIDIDR_EL1 with virtual domain counts KVM: arm64: Conditionally expose ID_AA64MMFR4_EL1.TLBID to guest KVM: arm64: Enable TLBIDIDR_EL1 access trap in guest KVM: arm64: Add tracepoints for TLBID vDomain load events Documentation/virt/kvm/api.rst | 58 +++ arch/arm64/include/asm/kvm_host.h | 53 +++ arch/arm64/include/uapi/asm/kvm.h | 2 + arch/arm64/kvm/Kconfig | 12 + arch/arm64/kvm/arm.c | 530 ++++++++++++++++++++++++++- arch/arm64/kvm/hisilicon/hisi_virt.c | 9 + arch/arm64/kvm/sys_regs.c | 76 +++- arch/arm64/kvm/sys_regs.h | 6 + arch/arm64/kvm/trace_arm.h | 47 +++ arch/arm64/tools/sysreg | 23 +- include/uapi/linux/kvm.h | 14 + 11 files changed, 826 insertions(+), 4 deletions(-) -- 2.33.0

2 9

[PATCH OLK-6.6] Add NI_RANDOM
by Luo Gengkun 14 Oct '26

14 Oct '26

Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> --- ...ed-fair-Proportional-newidle-balance.patch | 208 ++++++++++++++++++ ...complex-proportional-newidle-balance.patch | 133 +++++++++++ 2 files changed, 341 insertions(+) create mode 100644 NI/0001-sched-fair-Proportional-newidle-balance.patch create mode 100644 NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch diff --git a/NI/0001-sched-fair-Proportional-newidle-balance.patch b/NI/0001-sched-fair-Proportional-newidle-balance.patch new file mode 100644 index 000000000000..09ab81d809ff --- /dev/null +++ b/NI/0001-sched-fair-Proportional-newidle-balance.patch @@ -0,0 +1,208 @@ +From ccfc9b017997ffc44398363c85721a67c56533df Mon Sep 17 00:00:00 2001 +From: Peter Zijlstra <peterz(a)infradead.org> +Date: Fri, 7 Nov 2025 17:01:31 +0100 +Subject: [PATCH OLK-6.6 1/2] sched/fair: Proportional newidle balance + +Add a randomized algorithm that runs newidle balancing proportional to +its success rate. + +This improves schbench significantly: + + 6.18-rc4: 2.22 Mrps/s + 6.18-rc4+revert: 2.04 Mrps/s + 6.18-rc4+revert+random: 2.18 Mrps/S + +Conversely, per Adam Li this affects SpecJBB slightly, reducing it by 1%: + + 6.17: -6% + 6.17+revert: 0% + 6.17+revert+random: -1% + +Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org> +Reviewed-by: Dietmar Eggemann <dietmar.eggemann(a)arm.com> +Tested-by: Dietmar Eggemann <dietmar.eggemann(a)arm.com> +Tested-by: Chris Mason <clm(a)meta.com> +Link: https://lkml.kernel.org/r/6825c50d-7fa7-45d8-9b81-c6e7e25738e2@meta.com +Link: https://patch.msgid.link/20251107161739.770122091@infradead.org +Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> +--- + include/linux/sched/topology.h | 3 +++ + kernel/sched/core.c | 3 +++ + kernel/sched/fair.c | 43 ++++++++++++++++++++++++++++++---- + kernel/sched/features.h | 5 ++++ + kernel/sched/sched.h | 7 ++++++ + kernel/sched/topology.c | 6 +++++ + 6 files changed, 63 insertions(+), 4 deletions(-) + +diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h +index f0caa9e1a3c2..ab45a04061fb 100644 +--- a/include/linux/sched/topology.h ++++ b/include/linux/sched/topology.h +@@ -131,6 +131,9 @@ struct sched_domain { + unsigned int nr_balance_failed; /* initialise to 0 */ + + /* idle_balance() stats */ ++ unsigned int newidle_call; ++ unsigned int newidle_success; ++ unsigned int newidle_ratio; + u64 max_newidle_lb_cost; + unsigned long last_decay_max_lb_cost; + +diff --git a/kernel/sched/core.c b/kernel/sched/core.c +index 981652b0d14c..5b486736bd5d 100644 +--- a/kernel/sched/core.c ++++ b/kernel/sched/core.c +@@ -46,6 +46,7 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_se_tp); + EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp); + + DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); ++DEFINE_PER_CPU(struct rnd_state, sched_rnd_state); + + #ifdef CONFIG_SCHED_DEBUG + /* +@@ -9335,6 +9336,8 @@ void __init sched_init_smp(void) + { + sched_init_numa(); + ++ prandom_init_once(&sched_rnd_state); ++ + /* + * There's no userspace yet to cause hotplug operations; hence all the + * CPU masks are stable and all blatant races in the below code cannot +diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c +index fb01890861eb..31a1f7734332 100644 +--- a/kernel/sched/fair.c ++++ b/kernel/sched/fair.c +@@ -11840,8 +11840,23 @@ void update_max_interval(void) + max_load_balance_interval = HZ*num_online_cpus()/10; + } + +-static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost) ++static inline void update_newidle_stats(struct sched_domain *sd, unsigned int success) + { ++ sd->newidle_call++; ++ sd->newidle_success += success; ++ ++ if (sd->newidle_call >= 1024) { ++ sd->newidle_ratio = sd->newidle_success; ++ sd->newidle_call /= 2; ++ sd->newidle_success /= 2; ++ } ++} ++ ++static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost, unsigned int success) ++{ ++ if (cost) ++ update_newidle_stats(sd, success); ++ + if (cost > sd->max_newidle_lb_cost) { + /* + * Track max cost of a domain to make sure to not delay the +@@ -11889,7 +11904,7 @@ static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle) + * Decay the newidle max times here because this is a regular + * visit to all the domains. + */ +- need_decay = update_newidle_cost(sd, 0); ++ need_decay = update_newidle_cost(sd, 0, 0); + max_cost += sd->max_newidle_lb_cost; + + /* +@@ -12527,6 +12542,22 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + break; + + if (sd->flags & SD_BALANCE_NEWIDLE) { ++ unsigned int weight = 1; ++ ++ if (sched_feat(NI_RANDOM)) { ++ /* ++ * Throw a 1k sided dice; and only run ++ * newidle_balance according to the success ++ * rate. ++ */ ++ u32 d1k = sched_rng() % 1024; ++ weight = 1 + sd->newidle_ratio; ++ if (d1k > weight) { ++ update_newidle_stats(sd, 0); ++ continue; ++ } ++ weight = (1024 + weight/2) / weight; ++ } + + pulled_task = load_balance(this_cpu, this_rq, + sd, CPU_NEWLY_IDLE, +@@ -12534,10 +12565,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + + t1 = sched_clock_cpu(this_cpu); + domain_cost = t1 - t0; +- update_newidle_cost(sd, domain_cost); +- + curr_cost += domain_cost; + t0 = t1; ++ ++ /* ++ * Track max cost of a domain to make sure to not delay the ++ * next wakeup on the CPU. ++ */ ++ update_newidle_cost(sd, domain_cost, weight * !!pulled_task); + } + + /* +diff --git a/kernel/sched/features.h b/kernel/sched/features.h +index 578cd5ed2b93..0198f4e97d24 100644 +--- a/kernel/sched/features.h ++++ b/kernel/sched/features.h +@@ -117,3 +117,8 @@ SCHED_FEAT(STEAL, false) + SCHED_FEAT(SIS_NODE, false) + SCHED_FEAT(TEST_STEAL_PLUS, false) + SCHED_FEAT(TEST_STEAL_PLUS_IDLE, false) ++ ++/* ++ * Do newidle balancing proportional to its success rate using randomization. ++ */ ++SCHED_FEAT(NI_RANDOM, true) +\ No newline at end of file +diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h +index 8940656dac3d..1f5ea2538515 100644 +--- a/kernel/sched/sched.h ++++ b/kernel/sched/sched.h +@@ -4,6 +4,7 @@ + */ + #include <linux/sched.h> + ++#include <linux/prandom.h> + #include <linux/sched/autogroup.h> + #include <linux/sched/clock.h> + #include <linux/sched/coredump.h> +@@ -1196,6 +1197,12 @@ static inline bool is_migration_disabled(struct task_struct *p) + } + + DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); ++DECLARE_PER_CPU(struct rnd_state, sched_rnd_state); ++ ++static inline u32 sched_rng(void) ++{ ++ return prandom_u32_state(this_cpu_ptr(&sched_rnd_state)); ++} + + #define cpu_rq(cpu) (&per_cpu(runqueues, (cpu))) + #define this_rq() this_cpu_ptr(&runqueues) +diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c +index 6576bf380086..cb82bbc4fb41 100644 +--- a/kernel/sched/topology.c ++++ b/kernel/sched/topology.c +@@ -1822,6 +1822,12 @@ sd_init(struct sched_domain_topology_level *tl, + + .last_balance = jiffies, + .balance_interval = sd_weight, ++ ++ /* 50% success rate */ ++ .newidle_call = 512, ++ .newidle_success = 256, ++ .newidle_ratio = 512, ++ + .max_newidle_lb_cost = 0, + .last_decay_max_lb_cost = jiffies, + .child = child, +-- +2.34.1 + diff --git a/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch b/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch new file mode 100644 index 000000000000..4ab477726400 --- /dev/null +++ b/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch @@ -0,0 +1,133 @@ +From 069a09a4f06e2a05e933dc58bbeb8c66fcd75572 Mon Sep 17 00:00:00 2001 +From: Peter Zijlstra <peterz(a)infradead.org> +Date: Tue, 27 Jan 2026 16:17:48 +0100 +Subject: [PATCH OLK-6.6 2/2] sched/fair: More complex proportional newidle + balance + +It turns out that a few workloads (easyWave, fio) have a fairly low +success rate on newidle balance, but still benefit greatly from having +it anyway. + +Luckliky these workloads have a faily low newidle rate, so the cost if +doing the newidle is relatively low, even if unsuccessfull. + +Add a simple rate based part to the newidle ratio compute, such that +low rate newidle will still have a high newidle ratio. + +This cures the easyWave and fio workloads while not affecting the +schbench numbers either (which have a very high newidle rate). + +Reported-by: Mario Roy <marioeroy(a)gmail.com> +Reported-by: "Mohamed Abuelfotoh, Hazem" <abuehaze(a)amazon.com> +Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org> +Tested-by: Mario Roy <marioeroy(a)gmail.com> +Tested-by: "Mohamed Abuelfotoh, Hazem" <abuehaze(a)amazon.com> +Link: https://patch.msgid.link/20260127151748.GA1079264@noisy.programming.kicks-a… +Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> +--- + include/linux/sched/topology.h | 1 + + kernel/sched/fair.c | 27 +++++++++++++++++++++++++-- + kernel/sched/features.h | 3 ++- + kernel/sched/topology.c | 3 +++ + 4 files changed, 31 insertions(+), 3 deletions(-) + +diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h +index ab45a04061fb..d66e6c6c2cdf 100644 +--- a/include/linux/sched/topology.h ++++ b/include/linux/sched/topology.h +@@ -134,6 +134,7 @@ struct sched_domain { + unsigned int newidle_call; + unsigned int newidle_success; + unsigned int newidle_ratio; ++ u64 newidle_stamp; + u64 max_newidle_lb_cost; + unsigned long last_decay_max_lb_cost; + +diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c +index 31a1f7734332..ec40e79802bd 100644 +--- a/kernel/sched/fair.c ++++ b/kernel/sched/fair.c +@@ -11846,7 +11846,30 @@ static inline void update_newidle_stats(struct sched_domain *sd, unsigned int su + sd->newidle_success += success; + + if (sd->newidle_call >= 1024) { +- sd->newidle_ratio = sd->newidle_success; ++ u64 now = sched_clock(); ++ s64 delta = now - sd->newidle_stamp; ++ sd->newidle_stamp = now; ++ int ratio = 0; ++ ++ if (delta < 0) ++ delta = 0; ++ ++ if (sched_feat(NI_RATE)) { ++ /* ++ * ratio delta freq ++ * ++ * 1024 - 4 s - 128 Hz ++ * 512 - 2 s - 256 Hz ++ * 256 - 1 s - 512 Hz ++ * 128 - .5 s - 1024 Hz ++ * 64 - .25 s - 2048 Hz ++ */ ++ ratio = delta >> 22; ++ } ++ ++ ratio += sd->newidle_success; ++ ++ sd->newidle_ratio = min(1024, ratio); + sd->newidle_call /= 2; + sd->newidle_success /= 2; + } +@@ -12544,7 +12567,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + if (sd->flags & SD_BALANCE_NEWIDLE) { + unsigned int weight = 1; + +- if (sched_feat(NI_RANDOM)) { ++ if (sched_feat(NI_RANDOM) && sd->newidle_ratio < 1024) { + /* + * Throw a 1k sided dice; and only run + * newidle_balance according to the success +diff --git a/kernel/sched/features.h b/kernel/sched/features.h +index 0198f4e97d24..9a8b6f41f4cb 100644 +--- a/kernel/sched/features.h ++++ b/kernel/sched/features.h +@@ -121,4 +121,5 @@ SCHED_FEAT(TEST_STEAL_PLUS_IDLE, false) + /* + * Do newidle balancing proportional to its success rate using randomization. + */ +-SCHED_FEAT(NI_RANDOM, true) +\ No newline at end of file ++SCHED_FEAT(NI_RANDOM, true) ++SCHED_FEAT(NI_RATE, true) +diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c +index cb82bbc4fb41..ac198e940978 100644 +--- a/kernel/sched/topology.c ++++ b/kernel/sched/topology.c +@@ -4,6 +4,7 @@ + */ + #include "sched.h" + #include "sparsemask.h" ++#include <linux/sched/clock.h> + + DEFINE_MUTEX(sched_domains_mutex); + +@@ -1783,6 +1784,7 @@ sd_init(struct sched_domain_topology_level *tl, + struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu); + int sd_id, sd_weight, sd_flags = 0; + struct cpumask *sd_span; ++ u64 now = sched_clock(); + + #ifdef CONFIG_NUMA + /* +@@ -1827,6 +1829,7 @@ sd_init(struct sched_domain_topology_level *tl, + .newidle_call = 512, + .newidle_success = 256, + .newidle_ratio = 512, ++ .newidle_stamp = now, + + .max_newidle_lb_cost = 0, + .last_decay_max_lb_cost = jiffies, +-- +2.34.1 + -- 2.34.1

2 1

[PATCH OLK-6.6] Add NI_RANDOM
by Luo Gengkun 14 Oct '26

14 Oct '26

Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> --- NI/0000-cover-letter.patch | 21 ++ ...ed-fair-Proportional-newidle-balance.patch | 208 ++++++++++++++++++ ...complex-proportional-newidle-balance.patch | 133 +++++++++++ 3 files changed, 362 insertions(+) create mode 100644 NI/0000-cover-letter.patch create mode 100644 NI/0001-sched-fair-Proportional-newidle-balance.patch create mode 100644 NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch diff --git a/NI/0000-cover-letter.patch b/NI/0000-cover-letter.patch new file mode 100644 index 000000000000..b66e2ead869c --- /dev/null +++ b/NI/0000-cover-letter.patch @@ -0,0 +1,21 @@ +From 069a09a4f06e2a05e933dc58bbeb8c66fcd75572 Mon Sep 17 00:00:00 2001 +From: Luo Gengkun <luogengkun2(a)huawei.com> +Date: Wed, 14 Oct 2026 01:35:24 +0000 +Subject: [PATCH OLK-6.6 0/2] Backport NI_RANDOM for 5.15 + + +Peter Zijlstra (2): + sched/fair: Proportional newidle balance + sched/fair: More complex proportional newidle balance + + include/linux/sched/topology.h | 4 +++ + kernel/sched/core.c | 3 ++ + kernel/sched/fair.c | 66 +++++++++++++++++++++++++++++++--- + kernel/sched/features.h | 6 ++++ + kernel/sched/sched.h | 7 ++++ + kernel/sched/topology.c | 9 +++++ + 6 files changed, 91 insertions(+), 4 deletions(-) + +-- +2.34.1 + diff --git a/NI/0001-sched-fair-Proportional-newidle-balance.patch b/NI/0001-sched-fair-Proportional-newidle-balance.patch new file mode 100644 index 000000000000..09ab81d809ff --- /dev/null +++ b/NI/0001-sched-fair-Proportional-newidle-balance.patch @@ -0,0 +1,208 @@ +From ccfc9b017997ffc44398363c85721a67c56533df Mon Sep 17 00:00:00 2001 +From: Peter Zijlstra <peterz(a)infradead.org> +Date: Fri, 7 Nov 2025 17:01:31 +0100 +Subject: [PATCH OLK-6.6 1/2] sched/fair: Proportional newidle balance + +Add a randomized algorithm that runs newidle balancing proportional to +its success rate. + +This improves schbench significantly: + + 6.18-rc4: 2.22 Mrps/s + 6.18-rc4+revert: 2.04 Mrps/s + 6.18-rc4+revert+random: 2.18 Mrps/S + +Conversely, per Adam Li this affects SpecJBB slightly, reducing it by 1%: + + 6.17: -6% + 6.17+revert: 0% + 6.17+revert+random: -1% + +Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org> +Reviewed-by: Dietmar Eggemann <dietmar.eggemann(a)arm.com> +Tested-by: Dietmar Eggemann <dietmar.eggemann(a)arm.com> +Tested-by: Chris Mason <clm(a)meta.com> +Link: https://lkml.kernel.org/r/6825c50d-7fa7-45d8-9b81-c6e7e25738e2@meta.com +Link: https://patch.msgid.link/20251107161739.770122091@infradead.org +Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> +--- + include/linux/sched/topology.h | 3 +++ + kernel/sched/core.c | 3 +++ + kernel/sched/fair.c | 43 ++++++++++++++++++++++++++++++---- + kernel/sched/features.h | 5 ++++ + kernel/sched/sched.h | 7 ++++++ + kernel/sched/topology.c | 6 +++++ + 6 files changed, 63 insertions(+), 4 deletions(-) + +diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h +index f0caa9e1a3c2..ab45a04061fb 100644 +--- a/include/linux/sched/topology.h ++++ b/include/linux/sched/topology.h +@@ -131,6 +131,9 @@ struct sched_domain { + unsigned int nr_balance_failed; /* initialise to 0 */ + + /* idle_balance() stats */ ++ unsigned int newidle_call; ++ unsigned int newidle_success; ++ unsigned int newidle_ratio; + u64 max_newidle_lb_cost; + unsigned long last_decay_max_lb_cost; + +diff --git a/kernel/sched/core.c b/kernel/sched/core.c +index 981652b0d14c..5b486736bd5d 100644 +--- a/kernel/sched/core.c ++++ b/kernel/sched/core.c +@@ -46,6 +46,7 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_est_se_tp); + EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp); + + DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); ++DEFINE_PER_CPU(struct rnd_state, sched_rnd_state); + + #ifdef CONFIG_SCHED_DEBUG + /* +@@ -9335,6 +9336,8 @@ void __init sched_init_smp(void) + { + sched_init_numa(); + ++ prandom_init_once(&sched_rnd_state); ++ + /* + * There's no userspace yet to cause hotplug operations; hence all the + * CPU masks are stable and all blatant races in the below code cannot +diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c +index fb01890861eb..31a1f7734332 100644 +--- a/kernel/sched/fair.c ++++ b/kernel/sched/fair.c +@@ -11840,8 +11840,23 @@ void update_max_interval(void) + max_load_balance_interval = HZ*num_online_cpus()/10; + } + +-static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost) ++static inline void update_newidle_stats(struct sched_domain *sd, unsigned int success) + { ++ sd->newidle_call++; ++ sd->newidle_success += success; ++ ++ if (sd->newidle_call >= 1024) { ++ sd->newidle_ratio = sd->newidle_success; ++ sd->newidle_call /= 2; ++ sd->newidle_success /= 2; ++ } ++} ++ ++static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost, unsigned int success) ++{ ++ if (cost) ++ update_newidle_stats(sd, success); ++ + if (cost > sd->max_newidle_lb_cost) { + /* + * Track max cost of a domain to make sure to not delay the +@@ -11889,7 +11904,7 @@ static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle) + * Decay the newidle max times here because this is a regular + * visit to all the domains. + */ +- need_decay = update_newidle_cost(sd, 0); ++ need_decay = update_newidle_cost(sd, 0, 0); + max_cost += sd->max_newidle_lb_cost; + + /* +@@ -12527,6 +12542,22 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + break; + + if (sd->flags & SD_BALANCE_NEWIDLE) { ++ unsigned int weight = 1; ++ ++ if (sched_feat(NI_RANDOM)) { ++ /* ++ * Throw a 1k sided dice; and only run ++ * newidle_balance according to the success ++ * rate. ++ */ ++ u32 d1k = sched_rng() % 1024; ++ weight = 1 + sd->newidle_ratio; ++ if (d1k > weight) { ++ update_newidle_stats(sd, 0); ++ continue; ++ } ++ weight = (1024 + weight/2) / weight; ++ } + + pulled_task = load_balance(this_cpu, this_rq, + sd, CPU_NEWLY_IDLE, +@@ -12534,10 +12565,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + + t1 = sched_clock_cpu(this_cpu); + domain_cost = t1 - t0; +- update_newidle_cost(sd, domain_cost); +- + curr_cost += domain_cost; + t0 = t1; ++ ++ /* ++ * Track max cost of a domain to make sure to not delay the ++ * next wakeup on the CPU. ++ */ ++ update_newidle_cost(sd, domain_cost, weight * !!pulled_task); + } + + /* +diff --git a/kernel/sched/features.h b/kernel/sched/features.h +index 578cd5ed2b93..0198f4e97d24 100644 +--- a/kernel/sched/features.h ++++ b/kernel/sched/features.h +@@ -117,3 +117,8 @@ SCHED_FEAT(STEAL, false) + SCHED_FEAT(SIS_NODE, false) + SCHED_FEAT(TEST_STEAL_PLUS, false) + SCHED_FEAT(TEST_STEAL_PLUS_IDLE, false) ++ ++/* ++ * Do newidle balancing proportional to its success rate using randomization. ++ */ ++SCHED_FEAT(NI_RANDOM, true) +\ No newline at end of file +diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h +index 8940656dac3d..1f5ea2538515 100644 +--- a/kernel/sched/sched.h ++++ b/kernel/sched/sched.h +@@ -4,6 +4,7 @@ + */ + #include <linux/sched.h> + ++#include <linux/prandom.h> + #include <linux/sched/autogroup.h> + #include <linux/sched/clock.h> + #include <linux/sched/coredump.h> +@@ -1196,6 +1197,12 @@ static inline bool is_migration_disabled(struct task_struct *p) + } + + DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues); ++DECLARE_PER_CPU(struct rnd_state, sched_rnd_state); ++ ++static inline u32 sched_rng(void) ++{ ++ return prandom_u32_state(this_cpu_ptr(&sched_rnd_state)); ++} + + #define cpu_rq(cpu) (&per_cpu(runqueues, (cpu))) + #define this_rq() this_cpu_ptr(&runqueues) +diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c +index 6576bf380086..cb82bbc4fb41 100644 +--- a/kernel/sched/topology.c ++++ b/kernel/sched/topology.c +@@ -1822,6 +1822,12 @@ sd_init(struct sched_domain_topology_level *tl, + + .last_balance = jiffies, + .balance_interval = sd_weight, ++ ++ /* 50% success rate */ ++ .newidle_call = 512, ++ .newidle_success = 256, ++ .newidle_ratio = 512, ++ + .max_newidle_lb_cost = 0, + .last_decay_max_lb_cost = jiffies, + .child = child, +-- +2.34.1 + diff --git a/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch b/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch new file mode 100644 index 000000000000..4ab477726400 --- /dev/null +++ b/NI/0002-sched-fair-More-complex-proportional-newidle-balance.patch @@ -0,0 +1,133 @@ +From 069a09a4f06e2a05e933dc58bbeb8c66fcd75572 Mon Sep 17 00:00:00 2001 +From: Peter Zijlstra <peterz(a)infradead.org> +Date: Tue, 27 Jan 2026 16:17:48 +0100 +Subject: [PATCH OLK-6.6 2/2] sched/fair: More complex proportional newidle + balance + +It turns out that a few workloads (easyWave, fio) have a fairly low +success rate on newidle balance, but still benefit greatly from having +it anyway. + +Luckliky these workloads have a faily low newidle rate, so the cost if +doing the newidle is relatively low, even if unsuccessfull. + +Add a simple rate based part to the newidle ratio compute, such that +low rate newidle will still have a high newidle ratio. + +This cures the easyWave and fio workloads while not affecting the +schbench numbers either (which have a very high newidle rate). + +Reported-by: Mario Roy <marioeroy(a)gmail.com> +Reported-by: "Mohamed Abuelfotoh, Hazem" <abuehaze(a)amazon.com> +Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org> +Tested-by: Mario Roy <marioeroy(a)gmail.com> +Tested-by: "Mohamed Abuelfotoh, Hazem" <abuehaze(a)amazon.com> +Link: https://patch.msgid.link/20260127151748.GA1079264@noisy.programming.kicks-a… +Signed-off-by: Luo Gengkun <luogengkun2(a)huawei.com> +--- + include/linux/sched/topology.h | 1 + + kernel/sched/fair.c | 27 +++++++++++++++++++++++++-- + kernel/sched/features.h | 3 ++- + kernel/sched/topology.c | 3 +++ + 4 files changed, 31 insertions(+), 3 deletions(-) + +diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h +index ab45a04061fb..d66e6c6c2cdf 100644 +--- a/include/linux/sched/topology.h ++++ b/include/linux/sched/topology.h +@@ -134,6 +134,7 @@ struct sched_domain { + unsigned int newidle_call; + unsigned int newidle_success; + unsigned int newidle_ratio; ++ u64 newidle_stamp; + u64 max_newidle_lb_cost; + unsigned long last_decay_max_lb_cost; + +diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c +index 31a1f7734332..ec40e79802bd 100644 +--- a/kernel/sched/fair.c ++++ b/kernel/sched/fair.c +@@ -11846,7 +11846,30 @@ static inline void update_newidle_stats(struct sched_domain *sd, unsigned int su + sd->newidle_success += success; + + if (sd->newidle_call >= 1024) { +- sd->newidle_ratio = sd->newidle_success; ++ u64 now = sched_clock(); ++ s64 delta = now - sd->newidle_stamp; ++ sd->newidle_stamp = now; ++ int ratio = 0; ++ ++ if (delta < 0) ++ delta = 0; ++ ++ if (sched_feat(NI_RATE)) { ++ /* ++ * ratio delta freq ++ * ++ * 1024 - 4 s - 128 Hz ++ * 512 - 2 s - 256 Hz ++ * 256 - 1 s - 512 Hz ++ * 128 - .5 s - 1024 Hz ++ * 64 - .25 s - 2048 Hz ++ */ ++ ratio = delta >> 22; ++ } ++ ++ ratio += sd->newidle_success; ++ ++ sd->newidle_ratio = min(1024, ratio); + sd->newidle_call /= 2; + sd->newidle_success /= 2; + } +@@ -12544,7 +12567,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) + if (sd->flags & SD_BALANCE_NEWIDLE) { + unsigned int weight = 1; + +- if (sched_feat(NI_RANDOM)) { ++ if (sched_feat(NI_RANDOM) && sd->newidle_ratio < 1024) { + /* + * Throw a 1k sided dice; and only run + * newidle_balance according to the success +diff --git a/kernel/sched/features.h b/kernel/sched/features.h +index 0198f4e97d24..9a8b6f41f4cb 100644 +--- a/kernel/sched/features.h ++++ b/kernel/sched/features.h +@@ -121,4 +121,5 @@ SCHED_FEAT(TEST_STEAL_PLUS_IDLE, false) + /* + * Do newidle balancing proportional to its success rate using randomization. + */ +-SCHED_FEAT(NI_RANDOM, true) +\ No newline at end of file ++SCHED_FEAT(NI_RANDOM, true) ++SCHED_FEAT(NI_RATE, true) +diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c +index cb82bbc4fb41..ac198e940978 100644 +--- a/kernel/sched/topology.c ++++ b/kernel/sched/topology.c +@@ -4,6 +4,7 @@ + */ + #include "sched.h" + #include "sparsemask.h" ++#include <linux/sched/clock.h> + + DEFINE_MUTEX(sched_domains_mutex); + +@@ -1783,6 +1784,7 @@ sd_init(struct sched_domain_topology_level *tl, + struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu); + int sd_id, sd_weight, sd_flags = 0; + struct cpumask *sd_span; ++ u64 now = sched_clock(); + + #ifdef CONFIG_NUMA + /* +@@ -1827,6 +1829,7 @@ sd_init(struct sched_domain_topology_level *tl, + .newidle_call = 512, + .newidle_success = 256, + .newidle_ratio = 512, ++ .newidle_stamp = now, + + .max_newidle_lb_cost = 0, + .last_decay_max_lb_cost = jiffies, +-- +2.34.1 + -- 2.34.1

2 1

[PATCH OLK-6.6 0/2] Backport NI_RANDOM for 5.15
by Luo Gengkun 14 Oct '26

14 Oct '26

Peter Zijlstra (2): sched/fair: Proportional newidle balance sched/fair: More complex proportional newidle balance include/linux/sched/topology.h | 4 +++ kernel/sched/core.c | 3 ++ kernel/sched/fair.c | 66 +++++++++++++++++++++++++++++++--- kernel/sched/features.h | 6 ++++ kernel/sched/sched.h | 7 ++++ kernel/sched/topology.c | 9 +++++ 6 files changed, 91 insertions(+), 4 deletions(-) -- 2.34.1

1 0