Re: udmabuf vmap failed (Revert "udmabuf: fix vmap_udmabuf error page set" can help)

[Huan Yang]

在 2025/3/11 14:40, Bingbu Cao 写道:
> [You don't often get email from bingbu.cao@xxxxxxxxxxxxxxx. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>
> Huan Yang and Vivek,
>
> I am trying to use udmabuf for my test, and I cannot vmap the udmabuf
> buffers now. vmap_pfn_apply() will report a warning to complain that
> the pfns are invalid.
> I dump the pfn numbers as below:
> [ 3365.399641] pg[0] pfn 1148695
> [ 3365.399642] pg[1] pfn 1145057
> [ 3365.399642] pg[2] pfn 1134070
> [ 3365.399643] pg[3] pfn 1148700
> [ 3365.399643] pg[4] pfn 1144871
> [ 3365.399643] pg[5] pfn 1408686
> [ 3365.399643] pg[6] pfn 1408683
> ...
> [ 3365.399660] WARNING: CPU: 3 PID: 2772 at mm/vmalloc.c:3489 vmap_pfn_apply+0xb7/0xd0
> [ 3365.399667] Modules linked in:...
> [ 3365.399750] CPU: 3 UID: 0 PID: 2772 Comm: drm-test Not tainted 6.13.0-rc2-rvp #845
> [ 3365.399752] Hardware name: Intel Corporation Client Platform/xxxx, BIOS xxxFWI1.R00.3221.D83.2408120121 08/12/2024
> [ 3365.399753] RIP: 0010:vmap_pfn_apply+0xb7/0xd0
> [ 3365.399755] Code: 5b 41 5c 41 5d 5d c3 cc cc cc cc 48 21 c3 eb d1 48 21 c3 48 23 3d 31 c0 26 02 eb c5 48 c7 c7 c4 3c 20 a8 e8 5b c0 d8 ff eb 8a <0f> 0b b8 ea ff ff ff 5b 41 5c 41 5d 5d c3 cc cc cc cc 0f 1f 80 00
> [ 3365.399756] RSP: 0018:ffffb9b50c32fad0 EFLAGS: 00010202
> [ 3365.399757] RAX: 0000000000000001 RBX: 0000000000118717 RCX: 0000000000000000
> [ 3365.399758] RDX: 0000000080000000 RSI: ffffb9b50c358000 RDI: 00000000ffffffff
> [ 3365.399758] RBP: ffffb9b50c32fae8 R08: ffffb9b50c32fbd0 R09: 0000000000000001
> [ 3365.399759] R10: ffff941602479288 R11: 0000000000000000 R12: ffffb9b50c32fbd0
> [ 3365.399759] R13: ffff941618665ac0 R14: ffffb9b50c358000 R15: ffff941618665ac8
> [ 3365.399760] FS:  00007ff9e9ddd740(0000) GS:ffff94196f780000(0000) knlGS:0000000000000000
> [ 3365.399760] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 3365.399761] CR2: 000055fda5dc69d9 CR3: 00000001544de003 CR4: 0000000000f72ef0
> [ 3365.399762] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [ 3365.399762] DR3: 0000000000000000 DR6: 00000000ffff07f0 DR7: 0000000000000400
> [ 3365.399763] PKRU: 55555554
> [ 3365.399763] Call Trace:
> [ 3365.399765]  <TASK>
> [ 3365.399769]  ? show_regs+0x6d/0x80
> [ 3365.399773]  ? __warn+0x97/0x160
> [ 3365.399777]  ? vmap_pfn_apply+0xb7/0xd0
> [ 3365.399777]  ? report_bug+0x1ec/0x240
> [ 3365.399782]  ? handle_bug+0x63/0xa0
> [ 3365.399784]  ? exc_invalid_op+0x1d/0x80
> [ 3365.399785]  ? asm_exc_invalid_op+0x1f/0x30
> [ 3365.399790]  ? vmap_pfn_apply+0xb7/0xd0
> [ 3365.399791]  __apply_to_page_range+0x522/0x8a0
> [ 3365.399794]  ? __pfx_vmap_pfn_apply+0x10/0x10
> [ 3365.399795]  apply_to_page_range+0x18/0x20
> [ 3365.399796]  vmap_pfn+0x77/0xd0
> [ 3365.399797]  vmap_udmabuf+0xc5/0x110
> [ 3365.399802]  dma_buf_vmap+0x96/0x130
>
> I did an experiment to revert 18d7de823b7150344d242c3677e65d68c5271b04,
> then I can vmap the pages. Could you help what's wrong with that?

Sorry for that, as I reviewed pfn_valid, that's someting wired:

/**
 * pfn_valid - check if there is a valid memory map entry for a PFN
 * @pfn: the page frame number to check
 *
 * Check if there is a valid memory map entry aka struct page for the @pfn.
 * Note, that availability of the memory map entry does not imply that
 * there is actual usable memory at that @pfn. The struct page may
 * represent a hole or an unusable page frame.
 *
 * Return: 1 for PFNs that have memory map entries and 0 otherwise
 */

So, if pfn valid, it's return 1, else 0. So mean, only 1 is a valid pfn. 
But vmap_pfn_apply in there:

static int vmap_pfn_apply(pte_t *pte, unsigned long addr, void *private)
{
    struct vmap_pfn_data *data = private;
    unsigned long pfn = data->pfns[data->idx];
    pte_t ptent;

    if (WARN_ON_ONCE(pfn_valid(pfn)))
        return -EINVAL;

    ptent = pte_mkspecial(pfn_pte(pfn, data->prot));
    set_pte_at(&init_mm, addr, pte, ptent);

    data->idx++;
    return 0;
}

Do it give a wrong check? maybe should fix by:

static int vmap_pfn_apply(pte_t *pte, unsigned long addr, void *private)
{
    struct vmap_pfn_data *data = private;
    unsigned long pfn = data->pfns[data->idx];
    pte_t ptent;

-    if (WARN_ON_ONCE(pfn_valid(pfn)))
+    if (WARN_ON_ONCE(!pfn_valid(pfn)))
        return -EINVAL;

    ptent = pte_mkspecial(pfn_pte(pfn, data->prot));
    set_pte_at(&init_mm, addr, pte, ptent);

    data->idx++;
    return 0;
}

Please help me check it, also, you can apply this and then check it.:)

> --
> Best regards,
> Bingbu Cao