From: Jia He <jia.he@xxxxxxxxxxxxxxxx> Subject: mm/ksm.c: ignore STABLE_FLAG of rmap_item->address in rmap_walk_ksm() In our armv8a server(QDF2400), I noticed lots of WARN_ON caused by PAGE_SIZE unaligned for rmap_item->address under memory pressure tests(start 20 guests and run memhog in the host). --------------------------begin-------------------------------------- [ 410.853828] WARNING: CPU: 4 PID: 4641 at arch/arm64/kvm/../../../virt/kvm/arm/mmu.c:1826 kvm_age_hva_handler+0xc0/0xc8 [ 410.864518] Modules linked in: vhost_net vhost tap xt_CHECKSUM ipt_MASQUERADE nf_nat_masquerade_ipv4 ip6t_rpfilter ipt_REJECT nf_reject_ipv4 ip6t_REJECT nf_reject_ipv6 xt_conntrack ip_set nfnetlink ebtable_nat ebtable_broute bridge stp llc ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw ebtable_filter ebtables ip6table_filter ip6_tables iptable_filter rpcrdma ib_isert iscsi_target_mod ib_iser libiscsi scsi_transport_iscsi ib_srpt target_core_mod ib_srp scsi_transport_srp ib_ipoib rdma_ucm ib_ucm ib_umad rdma_cm ib_cm iw_cm mlx5_ib vfat fat ib_uverbs dm_mirror dm_region_hash ib_core dm_log dm_mod crc32_ce ipmi_ssif sg nfsd [ 410.935101] auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c mlx5_core ixgbe mlxfw devlink mdio ahci_platform libahci_platform qcom_emac libahci hdma hdma_mgmt i2c_qup [ 410.951369] CPU: 4 PID: 4641 Comm: memhog Tainted: G W 4.17.0-rc3+ #8 [ 410.959104] Hardware name: <snip for confidential issues> [ 410.969791] pstate: 80400005 (Nzcv daif +PAN -UAO) [ 410.974575] pc : kvm_age_hva_handler+0xc0/0xc8 [ 410.979012] lr : handle_hva_to_gpa+0xa8/0xe0 [ 410.983274] sp : ffff801761553290 [ 410.986581] x29: ffff801761553290 x28: 0000000000000000 [ 410.991888] x27: 0000000000000002 x26: 0000000000000000 [ 410.997195] x25: ffff801765430058 x24: ffff0000080b5608 [ 411.002501] x23: 0000000000000000 x22: ffff8017ccb84000 [ 411.007807] x21: 0000000003ff0000 x20: ffff8017ccb84000 [ 411.013113] x19: 000000000000fe00 x18: ffff000008fb3c08 [ 411.018419] x17: 0000000000000000 x16: 0060001645820bd3 [ 411.023725] x15: ffff80176aacbc08 x14: 0000000000000000 [ 411.029031] x13: 0000000000000040 x12: 0000000000000228 [ 411.034337] x11: 0000000000000000 x10: 0000000000000000 [ 411.039643] x9 : 0000000000000010 x8 : 0000000000000004 [ 411.044949] x7 : 0000000000000000 x6 : 00008017f0770000 [ 411.050255] x5 : 0000fffda59f0200 x4 : 0000000000000000 [ 411.055561] x3 : 0000000000000000 x2 : 000000000000fe00 [ 411.060867] x1 : 0000000003ff0000 x0 : 0000000020000000 [ 411.066173] Call trace: [ 411.068614] kvm_age_hva_handler+0xc0/0xc8 [ 411.072703] handle_hva_to_gpa+0xa8/0xe0 [ 411.076619] kvm_age_hva+0x4c/0xe8 [ 411.080014] kvm_mmu_notifier_clear_flush_young+0x54/0x98 [ 411.085408] __mmu_notifier_clear_flush_young+0x6c/0xa0 [ 411.090627] page_referenced_one+0x154/0x1d8 [ 411.094890] rmap_walk_ksm+0x12c/0x1d0 [ 411.098632] rmap_walk+0x94/0xa0 [ 411.101854] page_referenced+0x194/0x1b0 [ 411.105770] shrink_page_list+0x674/0xc28 [ 411.109772] shrink_inactive_list+0x26c/0x5b8 [ 411.114122] shrink_node_memcg+0x35c/0x620 [ 411.118211] shrink_node+0x100/0x430 [ 411.121778] do_try_to_free_pages+0xe0/0x3a8 [ 411.126041] try_to_free_pages+0xe4/0x230 [ 411.130045] __alloc_pages_nodemask+0x564/0xdc0 [ 411.134569] alloc_pages_vma+0x90/0x228 [ 411.138398] do_anonymous_page+0xc8/0x4d0 [ 411.142400] __handle_mm_fault+0x4a0/0x508 [ 411.146489] handle_mm_fault+0xf8/0x1b0 [ 411.150321] do_page_fault+0x218/0x4b8 [ 411.154064] do_translation_fault+0x90/0xa0 [ 411.158239] do_mem_abort+0x68/0xf0 [ 411.161721] el0_da+0x24/0x28 ---------------------------end--------------------------------------- In rmap_walk_ksm, the rmap_item->address might still have the STABLE_FLAG, then the start and end in handle_hva_to_gpa might not be PAGE_SIZE aligned. Thus it will cause exceptions in handle_hva_to_gpa on arm64. This patch fixes it by ignoring (not removing) the low bits of address when doing rmap_walk_ksm. IMO, it should be backported to stable tree. the stom of WARN_ONs is very easy for me to reproduce. More than that, I watched a panic (not reproducible) as follows: [35380.805825] page:ffff7fe003742d80 count:-4871 mapcount:-2126053375 mapping: (null) index:0x0 [35380.815024] flags: 0x1fffc00000000000() [35380.818845] raw: 1fffc00000000000 0000000000000000 0000000000000000 ffffecf981470000 [35380.826569] raw: dead000000000100 dead000000000200 ffff8017c001c000 0000000000000000 [35380.834294] page dumped because: nonzero _refcount [35380.839069] Modules linked in: vhost_net vhost tap ebtable_filter ebtables ip6table_filter ip6_tables iptable_filter fcoe libfcoe libfc 8021q garp mrp stp llc scsi_transport_fc openvswitch nf_conntrack_ipv6 nf_nat_ipv6 nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_defrag_ipv6 nf_nat nf_conntrack vfat fat rpcrdma ib_isert iscsi_target_mod ib_iser libiscsi scsi_transport_iscsi ib_srpt target_core_mod ib_srp scsi_transport_srp ib_ipoib rdma_ucm ib_ucm ib_uverbs ib_umad rdma_cm ib_cm iw_cm mlx5_ib ib_core crc32_ce ipmi_ssif tpm_tis tpm_tis_core sg nfsd auth_rpcgss nfs_acl lockd grace sunrpc dm_multipath ip_tables xfs libcrc32c mlx5_core mlxfw devlink ahci_platform libahci_platform libahci qcom_emac sdhci_acpi sdhci hdma mmc_core hdma_mgmt i2c_qup dm_mirror dm_region_hash dm_log dm_mod [35380.908341] CPU: 29 PID: 18323 Comm: qemu-kvm Tainted: G W 4.14.15-5.hxt.aarch64 #1 [35380.917107] Hardware name: <snip for confidential issues> [35380.930909] Call trace: [35380.933345] [<ffff000008088f00>] dump_backtrace+0x0/0x22c [35380.938723] [<ffff000008089150>] show_stack+0x24/0x2c [35380.943759] [<ffff00000893c078>] dump_stack+0x8c/0xb0 [35380.948794] [<ffff00000820ab50>] bad_page+0xf4/0x154 [35380.953740] [<ffff000008211ce8>] free_pages_check_bad+0x90/0x9c [35380.959642] [<ffff00000820c430>] free_pcppages_bulk+0x464/0x518 [35380.965545] [<ffff00000820db98>] free_hot_cold_page+0x22c/0x300 [35380.971448] [<ffff0000082176fc>] __put_page+0x54/0x60 [35380.976484] [<ffff0000080b1164>] unmap_stage2_range+0x170/0x2b4 [35380.982385] [<ffff0000080b12d8>] kvm_unmap_hva_handler+0x30/0x40 [35380.988375] [<ffff0000080b0104>] handle_hva_to_gpa+0xb0/0xec [35380.994016] [<ffff0000080b2644>] kvm_unmap_hva_range+0x5c/0xd0 [35380.999833] [<ffff0000080a8054>] I even injected a fault on purpose in kvm_unmap_hva_range by seting size=size-0x200, the call trace is similar as above. So I thought the panic is similarly caused by the root cause of WARN_ON. Andrea said: : It looks a straightforward safe fix, on x86 hva_to_gfn_memslot would : zap those bits and hide the misalignment caused by the low metadata : bits being erroneously left set in the address, but the arm code : notices when that's the last page in the memslot and the hva_end is : getting aligned and the size is below one page. : : I think the problem triggers in the addr += PAGE_SIZE of : unmap_stage2_ptes that never matches end because end is aligned but : addr is not. : : } while (pte++, addr += PAGE_SIZE, addr != end); : : x86 again only works on hva_start/hva_end after converting it to : gfn_start/end and that being in pfn units the bits are zapped before : they risk to cause trouble. Jia He said: : I've tested by myself in arm64 server (QDF2400,46 cpus,96G mem) Without : this patch, the WARN_ON is very easy for reproducing. After this patch, I : have run the same benchmarch for a whole day without any WARN_ONs Link: http://lkml.kernel.org/r/1525403506-6750-1-git-send-email-hejianet@xxxxxxxxx Signed-off-by: Jia He <jia.he@xxxxxxxxxxxxxxxx> Reviewed-by: Andrea Arcangeli <aarcange@xxxxxxxxxx> Tested-by: Jia He <hejianet@xxxxxxxxx> Cc: Suzuki K Poulose <Suzuki.Poulose@xxxxxxx> Cc: Minchan Kim <minchan@xxxxxxxxxx> Cc: Claudio Imbrenda <imbrenda@xxxxxxxxxxxxxxxxxx> Cc: Arvind Yadav <arvind.yadav.cs@xxxxxxxxx> Cc: Mike Rapoport <rppt@xxxxxxxxxxxxxxxxxx> Cc: <stable@xxxxxxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/ksm.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff -puN mm/ksm.c~mm-ksm-ignore-stable_flag-of-rmap_item-address-in-rmap_walk_ksm mm/ksm.c --- a/mm/ksm.c~mm-ksm-ignore-stable_flag-of-rmap_item-address-in-rmap_walk_ksm +++ a/mm/ksm.c @@ -216,6 +216,8 @@ struct rmap_item { #define SEQNR_MASK 0x0ff /* low bits of unstable tree seqnr */ #define UNSTABLE_FLAG 0x100 /* is a node of the unstable tree */ #define STABLE_FLAG 0x200 /* is listed from the stable tree */ +#define KSM_FLAG_MASK (SEQNR_MASK|UNSTABLE_FLAG|STABLE_FLAG) + /* to mask all the flags */ /* The stable and unstable tree heads */ static struct rb_root one_stable_tree[1] = { RB_ROOT }; @@ -2598,10 +2600,15 @@ again: anon_vma_lock_read(anon_vma); anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root, 0, ULONG_MAX) { + unsigned long addr; + cond_resched(); vma = vmac->vma; - if (rmap_item->address < vma->vm_start || - rmap_item->address >= vma->vm_end) + + /* Ignore the stable/unstable/sqnr flags */ + addr = rmap_item->address & ~KSM_FLAG_MASK; + + if (addr < vma->vm_start || addr >= vma->vm_end) continue; /* * Initially we examine only the vma which covers this @@ -2615,8 +2622,7 @@ again: if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg)) continue; - if (!rwc->rmap_one(page, vma, - rmap_item->address, rwc->arg)) { + if (!rwc->rmap_one(page, vma, addr, rwc->arg)) { anon_vma_unlock_read(anon_vma); return; } _
