From: Uladzislau Rezki <urezki@gmail.com>
To: Pranjal Arya <pranjal.arya@oss.qualcomm.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
Uladzislau Rezki <urezki@gmail.com>,
"Liam R. Howlett" <liam@infradead.org>,
Alice Ryhl <aliceryhl@google.com>,
Andrew Ballance <andrewjballance@gmail.com>,
linux-arm-msm@vger.kernel.org, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, maple-tree@lists.infradead.org,
Lorenzo Stoakes <ljs@kernel.org>,
Pranjal Shrivastava <praan@google.com>,
Will Deacon <will@kernel.org>,
Suzuki K Poulose <Suzuki.Poulose@arm.com>,
Neil Armstrong <neil.armstrong@linaro.org>,
Mostafa Saleh <smostafa@google.com>,
Balbir Singh <balbirs@nvidia.com>,
Suren Baghdasaryan <surenb@google.com>,
Marco Elver <elver@google.com>,
Dmitry Vyukov <dvyukov@google.com>,
Alexander Potapenko <glider@google.com>,
Shuah Khan <shuah@kernel.org>, Dev Jain <dev.jain@arm.com>,
Brendan Jackman <jackmanb@google.com>,
Puranjay Mohan <puranjay@kernel.org>,
Santosh Shukla <santosh.shukla@amd.com>,
Wyes Karny <wkarny@gmail.com>,
Sudeep Holla <sudeep.holla@kernel.org>
Subject: Re: [PATCH RFC 00/12] mm/vmalloc: migrate vmap_area indexing from rb-tree to maple-tree
Date: Sun, 14 Jun 2026 08:58:55 +0200 [thread overview]
Message-ID: <ai5RL6QNKJJMo6Fl@pc636> (raw)
In-Reply-To: <20260613-vmalloc_maple-v1-0-0aa740bb944b@oss.qualcomm.com>
On Sat, Jun 13, 2026 at 10:49:42PM +0530, Pranjal Arya wrote:
> vmalloc's free/busy/lazy area tracking is one of the last remaining
> augmented-rb_tree consumers in the core mm allocators. The rest of
> mm/ has been gradually consolidating range-keyed indexing around
> maple_tree (notably the per-process VMA tree in mm/mmap.c), and
> the underlying reason is a structural mismatch between rb_tree and
> range tracking:
>
> - rb_tree is a binary tree with a single entry per node. A lookup
> walks log2(N) nodes via a pointer chase that touches log2(N)
> cache lines, with one comparison per node. Range queries are not
> native to rb_tree; vmalloc has historically maintained them by
> augmenting every node with a subtree_max_size field whose value
> has to be recomputed on every insert, erase, and rotation via a
> custom callback. That callback machinery is vmalloc-specific
> code that exists only to coax range semantics out of a
> binary-keyed tree.
> - maple_tree stores up to 16 ranges per node (fanout f=16), so a
> lookup walks ~log16(N) nodes in tightly-packed pivot/slot arrays
> that are far more cache-friendly. Range queries are first-class
> (mas_empty_area, mas_find, mas_erase), with no augmentation
> callback to maintain. RCU traversal and sentinel range storage
> (XA_ZERO_ENTRY) are part of the data structure's contract, not
> bolted on by the consumer.
>
> For the vmalloc allocator's hot paths this means shallower walks
> under the same N, fewer cache lines touched per lookup, and no
> custom augmented-callback machinery to maintain. Aligning vmalloc
> with the same range-tree direction the rest of mm/ has taken
> collapses the augmented gap walker to a single mas_empty_area()
> descent, retires the augmented rb_node from struct vmap_area
> (-24 bytes per object on 64-bit), and exposes the range,
> sentinel, and RCU primitives needed for a per-CPU range cache
> that the augmented rb_tree could not host cleanly.
>
> This series completes the vmalloc internal migration from rb-tree
> indexed tracking to maple-tree indexed tracking for free, busy, and
> lazy vmap_area range management.
>
> The series removes vmalloc's internal rb-tree dependencies and moves
> address indexing and range lookups/scans to maple-tree-backed paths,
> while keeping ordered list neighbour traversal where coalescing
> semantics require stable predecessor/successor behaviour.
>
> In addition to the direct rb -> maple migration, the series includes
> robustness and scalability refinements in the same code path:
>
> - deferred/lazy maple bring-up to avoid early-boot allocator hazards
> - maple-assisted ordered-list insertion for busy/lazy tracking
> - mas_preallocate / mas_store_prealloc fast path for common-case
> publish work, with a non-indexed retry queue that absorbs the
> rare publish-under-pressure case without leaking or panicking
> - single mas_store(NULL, ...) sub-range trim in va_clip() in place
> of an erase-and-restore pair when narrowing a free-area entry
> - single mas_erase() for the busy-tree find-then-unlink pair on the
> free path
> - consolidation of in-use ranges as a single authoritative index on
> the steady-state allocation hot path
> - list_head representation of the lazy-purge queue, since that queue
> is bulk-drained and has no address-keyed query
> - per-CPU bump-allocator overlay layered on top of the migrated
> indices for short-lived, page-aligned, common-case allocations
> (design and chunk-size derivation in the 0010-0012 commit
> messages)
> - explicit lock/serialisation behaviour preserved (no lock removed)
>
> Primary advantages
> ==================
>
> - struct vmap_area shrinks by 24 bytes on 64-bit layouts (72 -> 48),
> removing the embedded augmented rb_node and the subtree_max_size
> field that the rb-tree gap walk depended on. Verified with pahole
> on arm64.
> - maple_tree's per-node fanout (multiple pivots/slots per node)
> replaces a binary rb-tree descent for indexed lookups, giving a
> shallower walk for the same allocation count.
> - alloc-side gap finding moves from a recursive augmented-rb walk to
> mas_empty_area() over the in-use range index, returning the lowest
> matching gap in a single descent.
> - vfree of a chunk-resident allocation through the per-CPU overlay
> resolves addr -> vmap_area in O(1) via the chunk's back-pointer
> array, with a bounded fast-reject for addresses outside any
> reserved chunk; the maple-tree busy lookup remains the fallback.
> - correctness behaviour preserved: ordered list neighbour traversal
> for coalescing remains; the locking/serialisation model is
> unchanged; lockdep is silent across the test_vmalloc subtest sweep.
> - robustness in bring-up and high churn: deferred/lazy maple
> initialisation avoids early-boot allocator hazards; the retry
> queue keeps publish failures under GFP_NOWAIT pressure correct
> without leaking or panicking.
>
> Real-silicon validation
> =======================
>
> The series was tested on Qualcomm Snapdragon X1E80100.
> The patched kernel was booted on the device against an RB baseline
> image of the same kernel revision, and exercised through:
>
> - GFXBench, run for several hours of sustained graphics workload;
> the patched kernel ran clean throughout, with throughput matching
> the RB baseline within run-to-run noise.
> - boot-time module loading via the finit_module / kernel_read_file
> path that exercises the bump-allocator's indexed write loop;
> this path drove the patch 0012 hardening, and the patched kernel
> is UBSAN-clean here.
> - repeated insmod / rmmod cycles to soak the chunk install / drain
> paths under live workload.
>
> No kernel WARN, BUG, or UBSAN report was observed across the
> multi-hour soak.
>
I tried to do some testing using this series. See below the kernel
splat. I use test_vmalloc.sh to test/stress it.
<snip>
[ 50.661082] ------------[ cut here ]------------
[ 50.662455] WARNING: mm/vmalloc.c:545 at vmap_small_pages_range_noflush+0x569/0x6c0, CPU#12: vmalloc_test/41/646
[ 50.664685] Modules linked in:
[ 50.665454] CPU: 12 UID: 0 PID: 646 Comm: vmalloc_test/41 Tainted: G W 7.1.0-rc7+ #347 PREEMPT(full)
[ 50.667765] Tainted: [W]=WARN
[ 50.668492] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 50.670639] RIP: 0010:vmap_small_pages_range_noflush+0x569/0x6c0
[ 50.671974] Code: e9 2d 01 00 00 b8 01 00 00 00 48 85 c2 0f 84 2d 01 00 00 4c 89 e6 4c 89 f7 e8 d3 80 c3 ff e9 1d 01 00 00 89 c3 e9 dc fe ff ff <0f> 0b 65 66 f7 05 fc 3e 14 02 ff ff 75 b2 65 48 8b 15 d9 3e 14 02
[ 50.676003] RSP: 0018:ffffcd36cc22bbe0 EFLAGS: 00010286
[ 50.677198] RAX: 8000000938940163 RBX: 0000000000000000 RCX: ffff8a7e9dd36000
[ 50.678752] RDX: fffffcb13b3a8000 RSI: ffff8a6fc919e800 RDI: 0000000000000000
[ 50.680313] RBP: ffffcd36cc22bc90 R08: ffffce6240000000 R09: ffffce622c064000
[ 50.681895] R10: 0000000000000024 R11: 000000000000000a R12: ffff8a7e9dd36000
[ 50.683459] R13: ffffce622c000000 R14: ffffce622c000000 R15: ffff8a776700db00
[ 50.685059] FS: 0000000000000000(0000) GS:ffff8a7f1a012000(0000) knlGS:0000000000000000
[ 50.686881] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 50.688165] CR2: 0000000000000000 CR3: 0000000839c3c000 CR4: 00000000000006f0
[ 50.689773] Call Trace:
[ 50.690419] <TASK>
[ 50.690994] __vmap_pages_range_noflush+0xda/0x100
[ 50.692100] __vmalloc_node_range_noprof+0x3e2/0x980
[ 50.693269] ? long_busy_list_alloc_test+0x68/0xf0
[ 50.694375] __vmalloc_node_noprof+0x52/0x70
[ 50.695366] ? long_busy_list_alloc_test+0x68/0xf0
[ 50.696470] vmalloc_noprof+0x25/0x30
[ 50.697389] long_busy_list_alloc_test+0x68/0xf0
[ 50.698436] ? __pfx_long_busy_list_alloc_test+0x10/0x10
[ 50.699626] test_func+0x112/0x230
[ 50.700442] ? __pfx_test_func+0x10/0x10
[ 50.701373] kthread+0x10d/0x140
[ 50.702153] ? __pfx_kthread+0x10/0x10
[ 50.703039] ret_from_fork+0x3a1/0x430
[ 50.703925] ? __pfx_kthread+0x10/0x10
[ 50.704821] ret_from_fork_asm+0x1a/0x30
[ 50.705736] </TASK>
[ 50.706322] irq event stamp: 1124795
[ 50.707157] hardirqs last enabled at (1124805): [<ffffffffa1fe09b8>] __up_console_sem+0x68/0x80
[ 50.709082] hardirqs last disabled at (1124816): [<ffffffffa1fe099d>] __up_console_sem+0x4d/0x80
[ 50.711031] softirqs last enabled at (0): [<ffffffffa1f1ba57>] copy_process+0xd67/0x2570
[ 50.712874] softirqs last disabled at (0): [<0000000000000000>] 0x0
[ 50.714247] ---[ end trace 0000000000000000 ]---
[ 50.715336] warn_alloc: 15 callbacks suppressed
[ 50.715339] vmalloc_test/41: vmalloc error: size 409600, failed to map pages, mode:0xcc2(GFP_KERNEL|__GFP_HIGHMEM), nodemask=(null),cpuset=/,mems_allowed=0
[ 50.719421] CPU: 12 UID: 0 PID: 646 Comm: vmalloc_test/41 Tainted: G W 7.1.0-rc7+ #347 PREEMPT(full)
[ 50.719424] Tainted: [W]=WARN
[ 50.719425] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 50.719426] Call Trace:
[ 50.719427] <TASK>
[ 50.719428] dump_stack_lvl+0x72/0xa0
[ 50.719433] dump_stack+0x14/0x1a
[ 50.719435] warn_alloc+0x137/0x160
[ 50.719445] __vmalloc_node_range_noprof+0x8cd/0x980
[ 50.719450] ? long_busy_list_alloc_test+0x68/0xf0
[ 50.719457] __vmalloc_node_noprof+0x52/0x70
[ 50.719460] ? long_busy_list_alloc_test+0x68/0xf0
[ 50.719461] vmalloc_noprof+0x25/0x30
[ 50.719463] long_busy_list_alloc_test+0x68/0xf0
[ 50.719465] ? __pfx_long_busy_list_alloc_test+0x10/0x10
[ 50.719467] test_func+0x112/0x230
[ 50.719472] ? __pfx_test_func+0x10/0x10
[ 50.719474] kthread+0x10d/0x140
[ 50.719476] ? __pfx_kthread+0x10/0x10
[ 50.719479] ret_from_fork+0x3a1/0x430
[ 50.719482] ? __pfx_kthread+0x10/0x10
[ 50.719484] ret_from_fork_asm+0x1a/0x30
[ 50.719493] </TASK>
[ 50.719494] Mem-Info:
[ 50.743128] active_anon:0 inactive_anon:0 isolated_anon:0
[ 50.743128] active_file:0 inactive_file:0 isolated_file:0
[ 50.743128] unevictable:0 dirty:0 writeback:0
[ 50.743128] slab_reclaimable:928 slab_unreclaimable:1820126
[ 50.743128] mapped:0 shmem:0 pagetables:610948
[ 50.743128] sec_pagetables:0 bounce:0
[ 50.743128] kernel_misc_reclaimable:0
[ 50.743128] free:12765547 free_pcp:26676 free_cma:0
[ 50.751177] Node 0 active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:0kB writeback:0kB shmem:0kB shmem_thp:0kB shmem_pmdmapped:0kB anon_thp:0kB kernel_stack:15136kB pagetables:2444312kB sec_pagetables:0kB all_unreclaimable? no Balloon:0kB gpu_active:0kB gpu_reclaim:0kB
[ 50.757522] Node 0 DMA free:11264kB boost:0kB min:12kB low:24kB high:36kB reserved_highatomic:0KB free_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB zspages:0kB present:15992kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
[ 50.763396] lowmem_reserve[]: 0 2991 64259 64259 64259
[ 50.764526] Node 0 DMA32 free:346136kB boost:0kB min:3144kB low:6204kB high:9264kB reserved_highatomic:0KB free_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB zspages:0kB present:3129216kB managed:3063680kB mlocked:0kB bounce:0kB free_pcp:54788kB local_pcp:760kB free_cma:0kB
[ 50.770714] lowmem_reserve[]: 0 0 61267 61267 61267
[ 50.771803] Node 0 Normal free:50727200kB boost:2048kB min:66468kB low:129204kB high:191940kB reserved_highatomic:0KB free_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB zspages:0kB present:63963136kB managed:62737744kB mlocked:0kB bounce:0kB free_pcp:51984kB local_pcp:520kB free_cma:0kB
[ 50.778200] lowmem_reserve[]: 0 0 0 0 0
[ 50.779086] Node 0 DMA: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 1*1024kB (U) 1*2048kB (M) 2*4096kB (M) = 11264kB
[ 50.781534] Node 0 DMA32: 31*4kB (U) 28*8kB (U) 34*16kB (U) 38*32kB (U) 41*64kB (U) 48*128kB (UM) 366*256kB (UM) 15*512kB (UM) 1*1024kB (M) 3*2048kB (UM) 55*4096kB (M) = 344700kB
[ 50.788250] Node 0 Normal: 1282565*4kB (UM) 889838*8kB (UM) 149191*16kB (UME) 301234*32kB (UME) 240946*64kB (UME) 84458*128kB (UME) 775*256kB (UE) 32*512kB (U) 1*1024kB (U) 1*2048kB (U) 0*4096kB = 50724532kB
[ 50.792662] Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
[ 50.794471] 0 total pagecache pages
[ 50.795273] 0 pages in swap cache
[ 50.796063] Free swap = 0kB
[ 50.796759] Total swap = 0kB
[ 50.797457] 16777086 pages RAM
[ 50.798184] 0 pages HighMem/MovableOnly
[ 50.799061] 322890 pages reserved
[ 50.799851] 0 pages hwpoisoned
[ 50.800578] Memory cgroup min protection 0kB -- low protection 0kB
[ 51.818069] BUG: TASK stack guard page was hit at ffffcd36cc25fff8 (stack is ffffcd36cc260000..ffffcd36cc264000)
[ 51.818078] Oops: stack guard page: 0000 [#1] SMP NOPTI
[ 51.818084] CPU: 31 UID: 0 PID: 653 Comm: vmalloc_test/48 Tainted: G W 7.1.0-rc7+ #347 PREEMPT(full)
[ 51.818088] Tainted: [W]=WARN
[ 51.818090] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 51.818091] RIP: 0010:__lock_acquire+0x8/0x21d0
[ 51.818098] Code: c7 c6 78 9f 40 a3 67 48 0f b9 3a e9 4e ff ff ff 66 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 55 48 89 e5 41 57 41 56 <41> 55 41 54 53 48 83 ec 48 48 8b 45 10 8b 1d b9 ac f6 01 48 89 45
[ 51.818100] RSP: 0018:ffffcd36cc260000 EFLAGS: 00010002
[ 51.818103] RAX: 0000000000000046 RBX: 0000000000000000 RCX: 0000000000000002
[ 51.818104] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffffffa3e065c8
[ 51.818105] RBP: ffffcd36cc260010 R08: 0000000000000001 R09: 0000000000000001
[ 51.818106] R10: 0000000000000002 R11: 0000000000000000 R12: 0000000000000002
[ 51.818107] R13: 0000000000000001 R14: ffffffffa3e065c8 R15: 0000000000000000
[ 51.818110] FS: 0000000000000000(0000) GS:ffff8a7f1a992000(0000) knlGS:0000000000000000
[ 51.818111] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 51.818112] CR2: ffffcd36cc25fff8 CR3: 0000000839c3c000 CR4: 00000000000006f0
[ 51.818116] Call Trace:
[ 51.818117] <TASK>
[ 51.818119] lock_acquire+0xc9/0x2e0
[ 51.818122] ? __alloc_pages_slowpath.constprop.0+0xca/0x14e0
[ 51.818129] seqcount_lockdep_reader_access+0x7a/0xa0
[ 51.818131] ? __alloc_pages_slowpath.constprop.0+0xca/0x14e0
[ 51.818133] __alloc_pages_slowpath.constprop.0+0xca/0x14e0
[ 51.818139] __alloc_frozen_pages_noprof+0x361/0x3b0
[ 51.818144] alloc_pages_mpol+0xa8/0x170
[ 51.818147] alloc_frozen_pages_noprof+0x54/0x90
[ 51.818148] ___kmalloc_large_node+0xc2/0xe0
[ 51.818151] __kmalloc_large_node_noprof+0x2e/0x130
[ 51.818154] __kvmalloc_node_noprof+0x5c0/0x9a0
[ 51.818157] ? vmap_bump_refill+0x44/0x1100
[ 51.818162] vmap_bump_refill+0x44/0x1100
[ 51.818164] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818166] ? lock_release+0xd1/0x2b0
[ 51.818169] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818171] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818173] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818176] alloc_vmap_area+0x1780/0x20b0
[ 51.818178] ? __lock_acquire+0x46d/0x21d0
[ 51.818181] ? lock_acquire+0xc9/0x2e0
[ 51.818182] ? find_held_lock+0x31/0x90
[ 51.818185] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818189] ? lock_release+0xd1/0x2b0
[ 51.818191] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818205] __get_vm_area_node+0xd1/0x130
[ 51.818209] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818210] ? vmap_bump_refill+0x44/0x1100
[ 51.818214] ? vmap_bump_refill+0x44/0x1100
[ 51.818217] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818220] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818222] ? vmap_bump_refill+0x44/0x1100
[ 51.818225] ? vmap_bump_refill+0x44/0x1100
[ 51.818228] vmap_bump_refill+0x44/0x1100
[ 51.818230] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818232] ? lock_release+0xd1/0x2b0
[ 51.818234] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818236] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818238] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818242] alloc_vmap_area+0x1780/0x20b0
[ 51.818243] ? __lock_acquire+0x46d/0x21d0
[ 51.818246] ? lock_acquire+0xc9/0x2e0
[ 51.818247] ? find_held_lock+0x31/0x90
[ 51.818249] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818252] ? lock_release+0xd1/0x2b0
[ 51.818254] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818258] __get_vm_area_node+0xd1/0x130
[ 51.818260] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818262] ? vmap_bump_refill+0x44/0x1100
[ 51.818265] ? vmap_bump_refill+0x44/0x1100
[ 51.818268] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818271] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818273] ? vmap_bump_refill+0x44/0x1100
[ 51.818276] ? vmap_bump_refill+0x44/0x1100
[ 51.818279] vmap_bump_refill+0x44/0x1100
[ 51.818281] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818283] ? lock_release+0xd1/0x2b0
[ 51.818285] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818287] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818289] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818293] alloc_vmap_area+0x1780/0x20b0
[ 51.818294] ? __lock_acquire+0x46d/0x21d0
[ 51.818297] ? lock_acquire+0xc9/0x2e0
[ 51.818298] ? find_held_lock+0x31/0x90
[ 51.818300] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818303] ? lock_release+0xd1/0x2b0
[ 51.818305] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818308] __get_vm_area_node+0xd1/0x130
[ 51.818311] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818312] ? vmap_bump_refill+0x44/0x1100
[ 51.818315] ? vmap_bump_refill+0x44/0x1100
[ 51.818318] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818321] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818323] ? vmap_bump_refill+0x44/0x1100
[ 51.818325] ? vmap_bump_refill+0x44/0x1100
[ 51.818329] vmap_bump_refill+0x44/0x1100
[ 51.818331] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818333] ? lock_release+0xd1/0x2b0
[ 51.818335] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818338] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818340] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818344] alloc_vmap_area+0x1780/0x20b0
[ 51.818345] ? __lock_acquire+0x46d/0x21d0
[ 51.818348] ? lock_acquire+0xc9/0x2e0
[ 51.818350] ? find_held_lock+0x31/0x90
[ 51.818352] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818354] ? lock_release+0xd1/0x2b0
[ 51.818357] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818360] __get_vm_area_node+0xd1/0x130
[ 51.818362] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818364] ? vmap_bump_refill+0x44/0x1100
[ 51.818367] ? vmap_bump_refill+0x44/0x1100
[ 51.818370] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818373] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818375] ? vmap_bump_refill+0x44/0x1100
[ 51.818377] ? vmap_bump_refill+0x44/0x1100
[ 51.818380] vmap_bump_refill+0x44/0x1100
[ 51.818382] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818385] ? lock_release+0xd1/0x2b0
[ 51.818386] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818389] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818390] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818394] alloc_vmap_area+0x1780/0x20b0
[ 51.818395] ? __lock_acquire+0x46d/0x21d0
[ 51.818398] ? lock_acquire+0xc9/0x2e0
[ 51.818400] ? find_held_lock+0x31/0x90
[ 51.818402] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818404] ? lock_release+0xd1/0x2b0
[ 51.818407] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818410] __get_vm_area_node+0xd1/0x130
[ 51.818412] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818414] ? vmap_bump_refill+0x44/0x1100
[ 51.818417] ? vmap_bump_refill+0x44/0x1100
[ 51.818420] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818423] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818425] ? vmap_bump_refill+0x44/0x1100
[ 51.818427] ? vmap_bump_refill+0x44/0x1100
[ 51.818430] vmap_bump_refill+0x44/0x1100
[ 51.818433] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818435] ? lock_release+0xd1/0x2b0
[ 51.818437] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818439] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818441] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818444] alloc_vmap_area+0x1780/0x20b0
[ 51.818445] ? __lock_acquire+0x46d/0x21d0
[ 51.818448] ? lock_acquire+0xc9/0x2e0
[ 51.818450] ? find_held_lock+0x31/0x90
[ 51.818452] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818454] ? lock_release+0xd1/0x2b0
[ 51.818456] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818459] __get_vm_area_node+0xd1/0x130
[ 51.818462] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818464] ? vmap_bump_refill+0x44/0x1100
[ 51.818467] ? vmap_bump_refill+0x44/0x1100
[ 51.818469] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818472] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818474] ? vmap_bump_refill+0x44/0x1100
[ 51.818477] ? vmap_bump_refill+0x44/0x1100
[ 51.818480] vmap_bump_refill+0x44/0x1100
[ 51.818482] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818484] ? lock_release+0xd1/0x2b0
[ 51.818486] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818488] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818490] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818493] alloc_vmap_area+0x1780/0x20b0
[ 51.818495] ? __lock_acquire+0x46d/0x21d0
[ 51.818497] ? lock_acquire+0xc9/0x2e0
[ 51.818499] ? find_held_lock+0x31/0x90
[ 51.818501] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818503] ? lock_release+0xd1/0x2b0
[ 51.818506] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818509] __get_vm_area_node+0xd1/0x130
[ 51.818511] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818513] ? vmap_bump_refill+0x44/0x1100
[ 51.818516] ? vmap_bump_refill+0x44/0x1100
[ 51.818519] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818521] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818524] ? vmap_bump_refill+0x44/0x1100
[ 51.818526] ? vmap_bump_refill+0x44/0x1100
[ 51.818529] vmap_bump_refill+0x44/0x1100
[ 51.818531] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818533] ? lock_release+0xd1/0x2b0
[ 51.818535] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818537] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818539] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818543] alloc_vmap_area+0x1780/0x20b0
[ 51.818544] ? __lock_acquire+0x46d/0x21d0
[ 51.818547] ? lock_acquire+0xc9/0x2e0
[ 51.818549] ? find_held_lock+0x31/0x90
[ 51.818550] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818553] ? lock_release+0xd1/0x2b0
[ 51.818555] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818558] __get_vm_area_node+0xd1/0x130
[ 51.818561] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818562] ? vmap_bump_refill+0x44/0x1100
[ 51.818565] ? vmap_bump_refill+0x44/0x1100
[ 51.818568] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818571] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818573] ? vmap_bump_refill+0x44/0x1100
[ 51.818575] ? vmap_bump_refill+0x44/0x1100
[ 51.818578] vmap_bump_refill+0x44/0x1100
[ 51.818581] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818583] ? lock_release+0xd1/0x2b0
[ 51.818585] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818587] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818589] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818592] alloc_vmap_area+0x1780/0x20b0
[ 51.818594] ? __lock_acquire+0x46d/0x21d0
[ 51.818596] ? lock_acquire+0xc9/0x2e0
[ 51.818598] ? find_held_lock+0x31/0x90
[ 51.818600] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818602] ? lock_release+0xd1/0x2b0
[ 51.818605] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818608] __get_vm_area_node+0xd1/0x130
[ 51.818610] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818612] ? vmap_bump_refill+0x44/0x1100
[ 51.818615] ? vmap_bump_refill+0x44/0x1100
[ 51.818618] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818620] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818623] ? vmap_bump_refill+0x44/0x1100
[ 51.818625] ? vmap_bump_refill+0x44/0x1100
[ 51.818628] vmap_bump_refill+0x44/0x1100
[ 51.818630] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818633] ? lock_release+0xd1/0x2b0
[ 51.818634] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818637] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818638] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818642] alloc_vmap_area+0x1780/0x20b0
[ 51.818643] ? __lock_acquire+0x46d/0x21d0
[ 51.818646] ? lock_acquire+0xc9/0x2e0
[ 51.818647] ? find_held_lock+0x31/0x90
[ 51.818649] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818652] ? lock_release+0xd1/0x2b0
[ 51.818654] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818657] __get_vm_area_node+0xd1/0x130
[ 51.818660] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818661] ? vmap_bump_refill+0x44/0x1100
[ 51.818664] ? vmap_bump_refill+0x44/0x1100
[ 51.818667] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818670] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818672] ? vmap_bump_refill+0x44/0x1100
[ 51.818674] ? vmap_bump_refill+0x44/0x1100
[ 51.818677] vmap_bump_refill+0x44/0x1100
[ 51.818679] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818682] ? lock_release+0xd1/0x2b0
[ 51.818684] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818686] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818688] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818691] alloc_vmap_area+0x1780/0x20b0
[ 51.818692] ? __lock_acquire+0x46d/0x21d0
[ 51.818695] ? lock_acquire+0xc9/0x2e0
[ 51.818697] ? find_held_lock+0x31/0x90
[ 51.818698] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818701] ? lock_release+0xd1/0x2b0
[ 51.818703] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818706] __get_vm_area_node+0xd1/0x130
[ 51.818709] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818710] ? vmap_bump_refill+0x44/0x1100
[ 51.818714] ? vmap_bump_refill+0x44/0x1100
[ 51.818717] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818720] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818722] ? vmap_bump_refill+0x44/0x1100
[ 51.818725] ? vmap_bump_refill+0x44/0x1100
[ 51.818728] vmap_bump_refill+0x44/0x1100
[ 51.818730] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818732] ? lock_release+0xd1/0x2b0
[ 51.818734] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818737] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818738] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818742] alloc_vmap_area+0x1780/0x20b0
[ 51.818743] ? __lock_acquire+0x46d/0x21d0
[ 51.818746] ? lock_acquire+0xc9/0x2e0
[ 51.818748] ? find_held_lock+0x31/0x90
[ 51.818750] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818753] ? lock_release+0xd1/0x2b0
[ 51.818755] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818758] __get_vm_area_node+0xd1/0x130
[ 51.818761] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818763] ? vmap_bump_refill+0x44/0x1100
[ 51.818766] ? vmap_bump_refill+0x44/0x1100
[ 51.818768] ? alloc_frozen_pages_noprof+0x54/0x90
[ 51.818771] __kvmalloc_node_noprof+0x2fd/0x9a0
[ 51.818774] ? vmap_bump_refill+0x44/0x1100
[ 51.818776] ? vmap_bump_refill+0x44/0x1100
[ 51.818779] vmap_bump_refill+0x44/0x1100
[ 51.818781] ? kmem_cache_free+0x2d3/0x4e0
[ 51.818784] ? lock_release+0xd1/0x2b0
[ 51.818786] ? vmap_bump_alloc+0x1ae/0x1e0
[ 51.818788] ? kmem_cache_free+0x2d8/0x4e0
[ 51.818790] ? vmap_bump_alloc+0x81/0x1e0
[ 51.818793] alloc_vmap_area+0x1780/0x20b0
[ 51.818795] ? __lock_acquire+0x46d/0x21d0
[ 51.818798] ? lock_acquire+0xc9/0x2e0
[ 51.818800] ? find_held_lock+0x31/0x90
[ 51.818802] ? __kmalloc_cache_node_noprof+0x18d/0x6e0
[ 51.818805] ? lock_release+0xd1/0x2b0
[ 51.818807] ? __kmalloc_cache_node_noprof+0x41a/0x6e0
[ 51.818810] __get_vm_area_node+0xd1/0x130
[ 51.818813] __vmalloc_node_range_noprof+0x145/0x980
[ 51.818814] ? long_busy_list_alloc_test+0x68/0xf0
[ 51.818818] ? kfree+0x32f/0x4a0
[ 51.818820] ? long_busy_list_alloc_test+0x68/0xf0
[ 51.818822] ? vfree+0xf6/0x2b0
[ 51.818825] __vmalloc_node_noprof+0x52/0x70
[ 51.818827] ? long_busy_list_alloc_test+0x68/0xf0
[ 51.818828] vmalloc_noprof+0x25/0x30
[ 51.818829] long_busy_list_alloc_test+0x68/0xf0
[ 51.818831] ? __pfx_long_busy_list_alloc_test+0x10/0x10
[ 51.818832] test_func+0x112/0x230
[ 51.818835] ? __pfx_test_func+0x10/0x10
[ 51.818836] kthread+0x10d/0x140
[ 51.818839] ? __pfx_kthread+0x10/0x10
[ 51.818842] ret_from_fork+0x3a1/0x430
[ 51.818845] ? __pfx_kthread+0x10/0x10
[ 51.818847] ret_from_fork_asm+0x1a/0x30
[ 51.818853] </TASK>
[ 51.818853] Modules linked in:
[ 51.818857] ---[ end trace 0000000000000000 ]---
[ 51.818859] RIP: 0010:__lock_acquire+0x8/0x21d0
[ 51.818860] Code: c7 c6 78 9f 40 a3 67 48 0f b9 3a e9 4e ff ff ff 66 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 55 48 89 e5 41 57 41 56 <41> 55 41 54 53 48 83 ec 48 48 8b 45 10 8b 1d b9 ac f6 01 48 89 45
[ 51.818861] RSP: 0018:ffffcd36cc260000 EFLAGS: 00010002
[ 51.818862] RAX: 0000000000000046 RBX: 0000000000000000 RCX: 0000000000000002
[ 51.818863] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffffffa3e065c8
[ 51.818864] RBP: ffffcd36cc260010 R08: 0000000000000001 R09: 0000000000000001
[ 51.818865] R10: 0000000000000002 R11: 0000000000000000 R12: 0000000000000002
[ 51.818865] R13: 0000000000000001 R14: ffffffffa3e065c8 R15: 0000000000000000
[ 51.818868] FS: 0000000000000000(0000) GS:ffff8a7f1a992000(0000) knlGS:0000000000000000
[ 51.818869] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 51.818869] CR2: ffffcd36cc25fff8 CR3: 0000000839c3c000 CR4: 00000000000006f0
[ 51.818873] Kernel panic - not syncing: Fatal exception in interrupt
[ 51.819725] Kernel Offset: 0x20c00000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
<snip>
I have not looked through the patches yet. See below some questions.
Do you solve any real issue by this series? Can we split this series
to several sections? I mean swap busy-tree then lazy-tree and latest
free-tree. This is for better code review reason and performance
evaluation one.
I am not sure we need to switch free-tree(it is not fragmented data structure)
whereas lazy and busy sounds good to me as those contain more objects.
--
Uladzislau Rezki
prev parent reply other threads:[~2026-06-14 6:59 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-13 17:19 [PATCH RFC 00/12] mm/vmalloc: migrate vmap_area indexing from rb-tree to maple-tree Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 01/12] mm/vmalloc: introduce maple_tree-based indexing for vmap_area Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 02/12] mm/vmalloc: convert allocation-side gap finding and insertion to maple_tree Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 03/12] mm/vmalloc: convert free, purge, and pcpu paths " Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 04/12] mm/vmalloc: finalize maple-only indexing and shrink struct vmap_area Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 05/12] mm/vmalloc: tighten failure handling under memory pressure Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 06/12] mm/vmalloc: tighten alloc/free hot paths Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 07/12] mm/vmalloc: consolidate occupied tree as authoritative index on hot path Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 08/12] mm/vmalloc: track lazy-purge queue as a list_head Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 09/12] mm/vmalloc: collapse busy-tree find-then-unlink into a single mas_erase Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 10/12] mm/vmalloc: per-CPU caching of free ranges from the maple_tree allocator Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 11/12] mm/vmalloc: O(1) lookup of cached vmap_areas with bounded fast-reject Pranjal Arya
2026-06-13 17:19 ` [PATCH RFC 12/12] mm/vmalloc: harden bump-allocator alloc/free against UBSAN array bounds Pranjal Arya
2026-06-13 23:15 ` [PATCH RFC 00/12] mm/vmalloc: migrate vmap_area indexing from rb-tree to maple-tree Matthew Wilcox
2026-06-14 6:35 ` [syzbot ci] " syzbot ci
2026-06-14 6:58 ` Uladzislau Rezki [this message]
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=ai5RL6QNKJJMo6Fl@pc636 \
--to=urezki@gmail.com \
--cc=Suzuki.Poulose@arm.com \
--cc=akpm@linux-foundation.org \
--cc=aliceryhl@google.com \
--cc=andrewjballance@gmail.com \
--cc=balbirs@nvidia.com \
--cc=dev.jain@arm.com \
--cc=dvyukov@google.com \
--cc=elver@google.com \
--cc=glider@google.com \
--cc=jackmanb@google.com \
--cc=liam@infradead.org \
--cc=linux-arm-msm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=ljs@kernel.org \
--cc=maple-tree@lists.infradead.org \
--cc=neil.armstrong@linaro.org \
--cc=praan@google.com \
--cc=pranjal.arya@oss.qualcomm.com \
--cc=puranjay@kernel.org \
--cc=santosh.shukla@amd.com \
--cc=shuah@kernel.org \
--cc=smostafa@google.com \
--cc=sudeep.holla@kernel.org \
--cc=surenb@google.com \
--cc=will@kernel.org \
--cc=wkarny@gmail.com \
/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