From: "Neronin, Niklas" <niklas.neronin@linux.intel.com>
To: "Michał Pecio" <michal.pecio@gmail.com>, mathias.nyman@linux.intel.com
Cc: linux-usb@vger.kernel.org
Subject: Re: [PATCH 5/4 RFC] An alternative dma_in_range() implementation
Date: Thu, 20 Feb 2025 15:18:08 +0200 [thread overview]
Message-ID: <bba69805-b28d-42f2-aa44-107a5700a8d3@linux.intel.com> (raw)
In-Reply-To: <20250220131451.6f356f31@foxbook>
On 20/02/2025 14.14, Michał Pecio wrote:
> Here's my attempt at it, the patch goes on top of the whole series. It
> compiles and passes basic testing (some URBs completed, some unlinked,
> no errors or other malfunction apparent).
>
> The idea is to translate dma to trb* ASAP and work with that, because
> all data structures use trb* so the code gets shorter and less verbose.
> This approach is also easy to adapt to changes in handle_tx_event(),
> for example it could trivially return trb* instead of seg* or take trb*
> (with or without accompanying seg*) and return bool or even seg*.
>
> I tried to make the common case (start_seg == end_seg) fast, both for
> hit and miss. 5 comparisons if the range wraps around, 4 if it doesn't.
>
Thanks, I have a similar version. If it's alright with you, I'll try to
combine your version and mine into one and add it to trb_in_td-v2.
Thanks,
Niklas
> diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
> index 600842425f6d..b18e7fd7d01e 100644
> --- a/drivers/usb/host/xhci-ring.c
> +++ b/drivers/usb/host/xhci-ring.c
> @@ -300,42 +300,36 @@ static struct xhci_segment *dma_in_range(struct xhci_segment *start_seg, union x
> dma_addr_t dma)
> {
> struct xhci_segment *seg = start_seg;
> + union xhci_trb *trb;
>
> - if (start_seg == end_seg) {
> - if (start_trb <= end_trb) {
> - if (xhci_trb_virt_to_dma(start_seg, start_trb) <= dma &&
> - dma <= xhci_trb_virt_to_dma(end_seg, end_trb))
> - return seg;
> - return NULL;
> + /* check if dma is a TRB in start_seg */
> + if (in_range(dma, seg->dma, TRB_SEGMENT_SIZE)) {
> + trb = seg->trbs + (dma - seg->dma) / sizeof(*trb);
> +
> + if (trb >= start_trb)
> + /* check if the range covers it and we are done */
> + return (end_seg != seg || trb <= end_trb) ? seg : NULL;
> +
> + /* check if the range circles back to the beginning of start_seg */
> + return (end_seg == seg && end_trb < start_trb && trb <= end_trb) ? seg : NULL;
> + }
> +
> + /* stop if the range doesn't pass through any other segment */
> + if (end_seg == seg && (end_trb >= start_trb || seg->next == seg))
> + return NULL;
> +
> + /* search remaining segments knowing that start_trb isn't there */
> + do {
> + seg = seg->next;
> +
> + if (in_range(dma, seg->dma, TRB_SEGMENT_SIZE)) {
> + trb = seg->trbs + (dma - seg->dma) / sizeof(*trb);
> +
> + return (seg != end_seg || trb <= end_trb) ? seg : NULL;
> }
> + } while (seg != end_seg && seg != start_seg);
>
> - /* Edge case, the TD wrapped around to the start segment. */
> - if (xhci_trb_virt_to_dma(end_seg, end_trb) < dma &&
> - dma < xhci_trb_virt_to_dma(start_seg, start_trb))
> - return NULL;
> - if (seg->dma <= dma && dma <= (seg->dma + TRB_SEGMENT_SIZE))
> - return seg;
> - seg = seg->next;
> - }
> -
> - /* Loop through segment which don't contain the DMA address. */
> - while (dma < seg->dma || (seg->dma + TRB_SEGMENT_SIZE) <= dma) {
> - if (seg == end_seg)
> - return NULL;
> -
> - seg = seg->next;
> - if (seg == start_seg)
> - return NULL;
> - }
> -
> - if (seg == start_seg) {
> - if (dma < xhci_trb_virt_to_dma(start_seg, start_trb))
> - return NULL;
> - } else if (seg == end_seg) {
> - if (xhci_trb_virt_to_dma(end_seg, end_trb) < dma)
> - return NULL;
> - }
> - return seg;
> + return NULL;
> }
>
> static struct xhci_segment *trb_in_td(struct xhci_td *td, dma_addr_t dma)
next prev parent reply other threads:[~2025-02-20 13:18 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-06 10:34 [PATCH 0/4] usb: xhci: improve trb_in_td() Niklas Neronin
2025-02-06 10:34 ` [PATCH 1/4] usb: xhci: refactor trb_in_td() to be static Niklas Neronin
2025-02-06 10:34 ` [PATCH 2/4] usb: xhci: move debug capabilities from trb_in_td() to handle_tx_event() Niklas Neronin
2025-03-05 8:46 ` Michał Pecio
2025-03-05 9:17 ` Neronin, Niklas
2025-02-06 10:34 ` [PATCH 3/4] usb: xhci: rework and simplify trb_in_td() Niklas Neronin
2025-02-19 8:56 ` Michał Pecio
2025-02-19 14:25 ` Mathias Nyman
2025-02-20 12:14 ` [PATCH 5/4 RFC] An alternative dma_in_range() implementation Michał Pecio
2025-02-20 13:18 ` Neronin, Niklas [this message]
2025-02-20 12:25 ` [PATCH 3/4] usb: xhci: rework and simplify trb_in_td() Neronin, Niklas
2025-02-06 10:34 ` [PATCH 4/4] usb: xhci: modify trb_in_td() to be more modular Niklas Neronin
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=bba69805-b28d-42f2-aa44-107a5700a8d3@linux.intel.com \
--to=niklas.neronin@linux.intel.com \
--cc=linux-usb@vger.kernel.org \
--cc=mathias.nyman@linux.intel.com \
--cc=michal.pecio@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