| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAO9qdTEXEZsA7rE8xJz6NLHTtCk0x4vWxCAd5_eXmPobbxEiNA@mail.gmail.com>
Date: Wed, 7 May 2025 22:07:08 +0900
From: Jeongjun Park <aha310510@...il.com>
To: Uladzislau Rezki <urezki@...il.com>
Cc: akpm@...ux-foundation.org, edumazet@...gle.com, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] mm/vmalloc: fix data race in show_numa_info()
Uladzislau Rezki <urezki@...il.com> wrote:
>
> On Tue, May 06, 2025 at 05:25:19PM +0900, Jeongjun Park wrote:
> > The following data-race was found in show_numa_info():
> >
> > ==================================================================
> > BUG: KCSAN: data-race in vmalloc_info_show / vmalloc_info_show
> >
> > read to 0xffff88800971fe30 of 4 bytes by task 8289 on cpu 0:
> > show_numa_info mm/vmalloc.c:4936 [inline]
> > vmalloc_info_show+0x5a8/0x7e0 mm/vmalloc.c:5016
> > seq_read_iter+0x373/0xb40 fs/seq_file.c:230
> > proc_reg_read_iter+0x11e/0x170 fs/proc/inode.c:299
> > new_sync_read fs/read_write.c:489 [inline]
> > vfs_read+0x5b4/0x740 fs/read_write.c:570
> > ksys_read+0xbe/0x190 fs/read_write.c:713
> > __do_sys_read fs/read_write.c:722 [inline]
> > __se_sys_read fs/read_write.c:720 [inline]
> > __x64_sys_read+0x41/0x50 fs/read_write.c:720
> > x64_sys_call+0x1729/0x1fd0 arch/x86/include/generated/asm/syscalls_64.h:1
> > do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> > do_syscall_64+0xa6/0x1b0 arch/x86/entry/syscall_64.c:94
> > entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >
> > write to 0xffff88800971fe30 of 4 bytes by task 8287 on cpu 1:
> > show_numa_info mm/vmalloc.c:4934 [inline]
> > vmalloc_info_show+0x38f/0x7e0 mm/vmalloc.c:5016
> > seq_read_iter+0x373/0xb40 fs/seq_file.c:230
> > proc_reg_read_iter+0x11e/0x170 fs/proc/inode.c:299
> > new_sync_read fs/read_write.c:489 [inline]
> > vfs_read+0x5b4/0x740 fs/read_write.c:570
> > ksys_read+0xbe/0x190 fs/read_write.c:713
> > __do_sys_read fs/read_write.c:722 [inline]
> > __se_sys_read fs/read_write.c:720 [inline]
> > __x64_sys_read+0x41/0x50 fs/read_write.c:720
> > x64_sys_call+0x1729/0x1fd0 arch/x86/include/generated/asm/syscalls_64.h:1
> > do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> > do_syscall_64+0xa6/0x1b0 arch/x86/entry/syscall_64.c:94
> > entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >
> > value changed: 0x0000008f -> 0x00000000
> > ==================================================================
> >
> > According to this report, there is a read/write data-race because m->private
> > is accessible to multiple CPUs. To fix this, instead of allocating the heap
> > in proc_vmalloc_init() and passing the heap address to m->private,
> > show_numa_info() should allocate the heap.
> >
> > One thing to note is that show_numa_info() is called in a critical section
> > of a spinlock, so it must be allocated on the heap with GFP_ATOMIC flag.
> >
> > Fixes: a47a126ad5ea ("vmallocinfo: add NUMA information")
> > Suggested-by: Eric Dumazet <edumazet@...gle.com>
> > Signed-off-by: Jeongjun Park <aha310510@...il.com>
> > ---
> > v2: Refactoring some functions and fix patch as per Eric Dumazet suggestion
> > - Link to v1: https://lore.kernel.org/all/20250505171948.24410-1-aha310510@gmail.com/
> > ---
> > mm/vmalloc.c | 50 +++++++++++++++++++++++++-------------------------
> > 1 file changed, 25 insertions(+), 25 deletions(-)
> >
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index 3ed720a787ec..a5e795346141 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -4914,28 +4914,32 @@ bool vmalloc_dump_obj(void *object)
> > #endif
> >
> > #ifdef CONFIG_PROC_FS
> > +
> > +/*
> > + * Print number of pages allocated on each memory node.
> > + *
> > + * This function can only be called if CONFIG_NUMA is enabled
> > + * and VM_UNINITIALIZED bit in v->flags is disabled.
> > + */
> > static void show_numa_info(struct seq_file *m, struct vm_struct *v)
> > {
> > - if (IS_ENABLED(CONFIG_NUMA)) {
> > - unsigned int nr, *counters = m->private;
> > - unsigned int step = 1U << vm_area_page_order(v);
> > + unsigned int nr, *counters;
> > + unsigned int step = 1U << vm_area_page_order(v);
> >
> > - if (!counters)
> > - return;
> > + counters = kcalloc(nr_node_ids, sizeof(unsigned int), GFP_ATOMIC);
> > + if (!counters)
> > + return;
> >
> > - if (v->flags & VM_UNINITIALIZED)
> > - return;
> > - /* Pair with smp_wmb() in clear_vm_uninitialized_flag() */
> > - smp_rmb();
> > + /* Pair with smp_wmb() in clear_vm_uninitialized_flag() */
> > + smp_rmb();
> >
> > - memset(counters, 0, nr_node_ids * sizeof(unsigned int));
> > + for (nr = 0; nr < v->nr_pages; nr += step)
> > + counters[page_to_nid(v->pages[nr])] += step;
> > + for_each_node_state(nr, N_HIGH_MEMORY)
> > + if (counters[nr])
> > + seq_printf(m, " N%u=%u", nr, counters[nr]);
> >
> > - for (nr = 0; nr < v->nr_pages; nr += step)
> > - counters[page_to_nid(v->pages[nr])] += step;
> > - for_each_node_state(nr, N_HIGH_MEMORY)
> > - if (counters[nr])
> > - seq_printf(m, " N%u=%u", nr, counters[nr]);
> > - }
> > + kfree(counters);
> > }
> >
> > static void show_purge_info(struct seq_file *m)
> > @@ -5013,7 +5017,10 @@ static int vmalloc_info_show(struct seq_file *m, void *p)
> > if (is_vmalloc_addr(v->pages))
> > seq_puts(m, " vpages");
> >
> > - show_numa_info(m, v);
> > + if (!(v->flags & VM_UNINITIALIZED) &&
> >
> I think it makes sense to move the VM_UNINITIALIZED check before:
>
> <snip>
> v = va->vm;
>
> seq_printf(m, "0x%pK-0x%pK %7ld",
> v->addr, v->addr + v->size, v->size);
> <snip>
>
> because it can be still un-initialized, thus flags like "nr_pages", etc
> will not be valid.
>
> Any thoughts? It has nothing to do with a patch, because it fixes other
> race issue and what i propose might well be a separate patch if we agree.
>
> Thanks!
>
> --
> Uladzislau Rezki
That's absolutely correct. If the VM_UNINITIALIZED bit is enabled, it
means there are uninitialized member variables, so printing the vm_struct
information is not very appropriate.
I think it would be better to add the VM_UNINITIALIZED bit check code
right after `v = va->vm;`.
This change isn't that complicated, so I'll add it to the v3 patch and
send it.
Regards,
Jeongjun Park
Powered by blists - more mailing lists