Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next

Pull networking updates from David Miller:

 1) Support IPV6 RA Captive Portal Identifier, from Maciej Żenczykowski.

 2) Use bio_vec in the networking instead of custom skb_frag_t, from
    Matthew Wilcox.

 3) Make use of xmit_more in r8169 driver, from Heiner Kallweit.

 4) Add devmap_hash to xdp, from Toke Høiland-Jørgensen.

 5) Support all variants of 5750X bnxt_en chips, from Michael Chan.

 6) More RTNL avoidance work in the core and mlx5 driver, from Vlad
    Buslov.

 7) Add TCP syn cookies bpf helper, from Petar Penkov.

 8) Add 'nettest' to selftests and use it, from David Ahern.

 9) Add extack support to drop_monitor, add packet alert mode and
    support for HW drops, from Ido Schimmel.

10) Add VLAN offload to stmmac, from Jose Abreu.

11) Lots of devm_platform_ioremap_resource() conversions, from
    YueHaibing.

12) Add IONIC driver, from Shannon Nelson.

13) Several kTLS cleanups, from Jakub Kicinski.

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1930 commits)
  mlxsw: spectrum_buffers: Add the ability to query the CPU port's shared buffer
  mlxsw: spectrum: Register CPU port with devlink
  mlxsw: spectrum_buffers: Prevent changing CPU port's configuration
  net: ena: fix incorrect update of intr_delay_resolution
  net: ena: fix retrieval of nonadaptive interrupt moderation intervals
  net: ena: fix update of interrupt moderation register
  net: ena: remove all old adaptive rx interrupt moderation code from ena_com
  net: ena: remove ena_restore_ethtool_params() and relevant fields
  net: ena: remove old adaptive interrupt moderation code from ena_netdev
  net: ena: remove code duplication in ena_com_update_nonadaptive_moderation_interval _*()
  net: ena: enable the interrupt_moderation in driver_supported_features
  net: ena: reimplement set/get_coalesce()
  net: ena: switch to dim algorithm for rx adaptive interrupt moderation
  net: ena: add intr_moder_rx_interval to struct ena_com_dev and use it
  net: phy: adin: implement Energy Detect Powerdown mode via phy-tunable
  ethtool: implement Energy Detect Powerdown support via phy-tunable
  xen-netfront: do not assume sk_buff_head list is empty in error handling
  s390/ctcm: Delete unnecessary checks before the macro call “dev_kfree_skb”
  net: ena: don't wake up tx queue when down
  drop_monitor: Better sanitize notified packets
  ...

1 files changed, 21 insertions, 6 deletions

diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
index 0a00eaca6fae..ddd8addcdb5c 100644
--- a/kernel/bpf/cgroup.c
+++ b/kernel/bpf/cgroup.c
@@ -964,7 +964,6 @@ static int sockopt_alloc_buf(struct bpf_sockopt_kern *ctx, int max_optlen)
 		return -ENOMEM;
 
 	ctx->optval_end = ctx->optval + max_optlen;
-	ctx->optlen = max_optlen;
 
 	return 0;
 }
@@ -984,7 +983,7 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level,
 		.level = *level,
 		.optname = *optname,
 	};
-	int ret;
+	int ret, max_optlen;
 
 	/* Opportunistic check to see whether we have any BPF program
 	 * attached to the hook so we don't waste time allocating
@@ -994,10 +993,18 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level,
 	    __cgroup_bpf_prog_array_is_empty(cgrp, BPF_CGROUP_SETSOCKOPT))
 		return 0;
 
-	ret = sockopt_alloc_buf(&ctx, *optlen);
+	/* Allocate a bit more than the initial user buffer for
+	 * BPF program. The canonical use case is overriding
+	 * TCP_CONGESTION(nv) to TCP_CONGESTION(cubic).
+	 */
+	max_optlen = max_t(int, 16, *optlen);
+
+	ret = sockopt_alloc_buf(&ctx, max_optlen);
 	if (ret)
 		return ret;
 
+	ctx.optlen = *optlen;
+
 	if (copy_from_user(ctx.optval, optval, *optlen) != 0) {
 		ret = -EFAULT;
 		goto out;
@@ -1016,7 +1023,7 @@ int __cgroup_bpf_run_filter_setsockopt(struct sock *sk, int *level,
 	if (ctx.optlen == -1) {
 		/* optlen set to -1, bypass kernel */
 		ret = 1;
-	} else if (ctx.optlen > *optlen || ctx.optlen < -1) {
+	} else if (ctx.optlen > max_optlen || ctx.optlen < -1) {
 		/* optlen is out of bounds */
 		ret = -EFAULT;
 	} else {
@@ -1063,6 +1070,8 @@ int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level,
 	if (ret)
 		return ret;
 
+	ctx.optlen = max_optlen;
+
 	if (!retval) {
 		/* If kernel getsockopt finished successfully,
 		 * copy whatever was returned to the user back
@@ -1325,6 +1334,7 @@ static u32 sysctl_convert_ctx_access(enum bpf_access_type type,
 				     struct bpf_prog *prog, u32 *target_size)
 {
 	struct bpf_insn *insn = insn_buf;
+	u32 read_size;
 
 	switch (si->off) {
 	case offsetof(struct bpf_sysctl, write):
@@ -1356,7 +1366,9 @@ static u32 sysctl_convert_ctx_access(enum bpf_access_type type,
 				treg, si->dst_reg,
 				offsetof(struct bpf_sysctl_kern, ppos));
 			*insn++ = BPF_STX_MEM(
-				BPF_SIZEOF(u32), treg, si->src_reg, 0);
+				BPF_SIZEOF(u32), treg, si->src_reg,
+				bpf_ctx_narrow_access_offset(
+					0, sizeof(u32), sizeof(loff_t)));
 			*insn++ = BPF_LDX_MEM(
 				BPF_DW, treg, si->dst_reg,
 				offsetof(struct bpf_sysctl_kern, tmp_reg));
@@ -1365,8 +1377,11 @@ static u32 sysctl_convert_ctx_access(enum bpf_access_type type,
 				BPF_FIELD_SIZEOF(struct bpf_sysctl_kern, ppos),
 				si->dst_reg, si->src_reg,
 				offsetof(struct bpf_sysctl_kern, ppos));
+			read_size = bpf_size_to_bytes(BPF_SIZE(si->code));
 			*insn++ = BPF_LDX_MEM(
-				BPF_SIZE(si->code), si->dst_reg, si->dst_reg, 0);
+				BPF_SIZE(si->code), si->dst_reg, si->dst_reg,
+				bpf_ctx_narrow_access_offset(
+					0, read_size, sizeof(loff_t)));
 		}
 		*target_size = sizeof(u32);
 		break;