Message ID | 20230726030936.1587269-1-trdgn@amazon.com (mailing list archive) |
---|---|
State | Superseded |
Delegated to: | Netdev Maintainers |
Headers | show |
Series | tun: avoid high-order page allocation for packet header | expand |
On Tue, 25 Jul 2023 20:09:36 -0700 Tahsin Erdogan wrote: > @@ -1838,6 +1838,9 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, > */ > zerocopy = false; > } else { > + if (linear == 0) > + linear = min_t(size_t, good_linear, copylen); nit: would you mind changing to !linear instead of linear == 0 ? Also - I don't see linear explicitly getting set to 0. What guarantees that? What's the story there? Otherwise seems reasonable. One more allocation but hopefully nobody will notice.
Thanks Jakub for considering this patch. On Mon, 2023-07-31 at 13:58 -0700, Jakub Kicinski wrote: > On Tue, 25 Jul 2023 20:09:36 -0700 Tahsin Erdogan wrote: > > @@ -1838,6 +1838,9 @@ static ssize_t tun_get_user(struct tun_struct > > *tun, struct tun_file *tfile, > > */ > > zerocopy = false; > > } else { > > + if (linear == 0) > > + linear = min_t(size_t, good_linear, > > copylen); > > nit: would you mind changing to !linear instead of linear == 0 ? Yes, I will fix this in v2. > Also - I don't see linear explicitly getting set to 0. What > guarantees > that? What's the story there? linear is set to 0 in the earlier if block. When gso.hdr_len is 0, linear also becomes 0: if (!zerocopy) { copylen = len; if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) linear = good_linear; else linear = tun16_to_cpu(tun, gso.hdr_len); }
diff --git a/drivers/net/tun.c b/drivers/net/tun.c index d75456adc62a..b9d111fbea63 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -1523,7 +1523,7 @@ static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, int err; /* Under a page? Don't bother with paged skb. */ - if (prepad + len < PAGE_SIZE || !linear) + if (prepad + len < PAGE_SIZE) linear = len; skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, @@ -1838,6 +1838,9 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, */ zerocopy = false; } else { + if (linear == 0) + linear = min_t(size_t, good_linear, copylen); + skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); }
When GSO is not enabled and a packet is transmitted via writev(), all payload is treated as header which requires a contiguous memory allocation. This allocation request is harder to satisfy, and may even fail if there is enough fragmentation. Note that sendmsg() code path limits the linear copy length, so this change makes writev() and sendmsg() more consistent. Signed-off-by: Tahsin Erdogan <trdgn@amazon.com> --- drivers/net/tun.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)