xfrm: set-mark default behavior changes
Subsystems (3)
cover xfrm: set-mark default behavior changes
Resend; added authors of 9b42c1f179a6 to CC list
A behavior change introduced in 9b42c1f179a6 (“xfrm: Extend the output_mark to support input direction and masking”) results in a change in:
1. Default outbound behavior with regards to route lookup marks, and
2. Inbound behavior for SAs used to decapsulate packets when the output
mark (as specified in 4.14 to 4.18) is set.
This patch set restores the previous default outbound behavior, resolving (1), but behavior change (2) will require more discussion.
Specifically, in (2), a SA with a "output mark" set will now have that Mark imposed on the inbound packet (As opposed to the previous output-mark behavior where the inbound packet's mark would not be touched). This is less of a concern, as it is limited to the case where:
1. SA output mark is set
2. SA is using non-transport mode
3. SA is configured for inbound decapsulation (local dst IP)
Critically, conditions 1 and 3 imply a configuration that output mark was not designed to support. The only valid use case for this seems to be the loopback case (as IP addresses would apply bidirectionally). As such, we believe that this behavioral change is acceptable as is.
Benedict Wong (1):
xfrm: Make set-mark default behavior backward compatible
net/xfrm/xfrm_policy.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
Signature · 1 line
2.20.1.97.g81188d93c3-goog
PATCH RESEND 1/1 xfrm: Make set-mark default behavior backward compatible
Fixes 9b42c1f179a6, which changed the default route lookup behavior for tunnel mode SAs in the outbound direction to use the skb mark, whereas previously mark=0 was used if the output mark was unspecified. In mark-based routing schemes such as Android’s, this change in default behavior causes routing loops or lookup failures.
This patch restores the default behavior of using a 0 mark while still incorporating the skb mark if the SET_MARK (and SET_MARK_MASK) is specified.
Tested with additions to Android's kernel unit test suite:
https://android-review.googlesource.com/c/kernel/tests/+/860150
--- net/xfrm/xfrm_policy.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
net/xfrm/xfrm_policy.c +4 -1
@@ -2600,7 +2600,10 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy,dst_copy_metrics (dst1, dst);if (xfrm[i]->props.mode !=XFRM_MODE_TRANSPORT ) { -__u32 mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]); +__u32 mark =0 ; + +if (xfrm[i]->props.smark.v || xfrm[i]->props.smark.m) + mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]); family = xfrm[i]->props.family; dst = xfrm_dst_lookup(xfrm[i], tos, fl->flowi_oif,
Signature · 1 line
2.20.1.97.g81188d93c3-goog
Re: [PATCH ipsec, resend 1/1] xfrm: Make set-mark default behavior backward compatible
Hi Benedict,
Quoted 9 lines · On Mon, 14 Jan 2019 11:24:38 -0800 Benedict Wong <[email protected]>…
Fixes
9b42c1f179a6, which changed the default route lookup behavior for tunnel mode SAs in the outbound direction to use the skb mark, whereas previously mark=0 was used if the output mark was unspecified. In mark-based routing schemes such as Android’s, this change in default behavior causes routing loops or lookup failures.This patch restores the default behavior of using a 0 mark while still incorporating the skb mark if the
SET_MARK(andSET_MARK_MASK) is specified.
It's a little sad that we didn't distinguish between 'no-mark-provided' and 'mark = mask = 0'. IIUC before this fix, explicitly setting mark, mask to a) mark = 0, mask = 0xffffffff and b) mark = 0, mask = 0 behaved differently, whereas after this fix they will act the same.
But as it seems there was never support for the (b) scenario and commit 9b42c1f179a6 broke the existing behavior, so I'm ok with this fix.
Thanks!
Eyal.
Quoted 29 lines
Tested with additions to Android's kernel unit test suite:
https://android-review.googlesource.com/c/kernel/tests/+/860150Fixes 9b42c1f179a6 ("xfrm: Extend the output_mark to support inputdirection and masking") Signed-off-by: Benedict Wong <[email protected]> ---
net/xfrm/xfrm_policy.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)net/xfrm/xfrm_policy.c +4
-1@@ -2600,7 +2600,10 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy, dst_copy_metrics(dst1, dst);if (xfrm[i]->props.mode !=XFRM_MODE_TRANSPORT ) { -__u32 mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]); +__u32 mark =0 ; + +if (xfrm[i]->props.smark.v || xfrm[i]->props.smark.m) + mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]);net/xfrm/xfrm_policy.c +0
-0family = xfrm[i]->props.family; dst = xfrm_dst_lookup(xfrm[i], tos,
fl->flowi_oif,
Re: [PATCH ipsec, resend 1/1] xfrm: Make set-mark default behavior backward compatible
Hi Eyal,
Thanks for taking a look at this.
Quoted 4 lines
It's a little sad that we didn't distinguish between 'no-mark-provided' and 'mark = mask = 0'. IIUC before this fix, explicitly setting mark, mask to a) mark = 0, mask =
0xffffffffand b) mark = 0, mask = 0 behaved differently, whereas after this fix they will act the same.
Agreed. It's a little hacky, but someone determined to use the entirety of the mark from the flowi (case b) in the route lookup could use mark = 0x1, mask = 0x0, and that would give the same results as mark = 0x0, mask = 0x0. It's a little intuitive though. :)
Perhaps at some future point, it may be worth defining a bit in the xfrm_state->props->extra_flags as "set-mark provided," At which point we can separate cases (a) and (b)
Cheers,
Ben.
Quoted 57 lines · Eyal Birger
Hi Benedict,
Quoted 9 lines · On Mon, 14 Jan 2019 11:24:38 -0800 Benedict Wong <[email protected]>…
Fixes
9b42c1f179a6, which changed the default route lookup behavior for tunnel mode SAs in the outbound direction to use the skb mark, whereas previously mark=0 was used if the output mark was unspecified. In mark-based routing schemes such as Android’s, this change in default behavior causes routing loops or lookup failures.This patch restores the default behavior of using a 0 mark while still incorporating the skb mark if the
SET_MARK(andSET_MARK_MASK) is specified.It's a little sad that we didn't distinguish between 'no-mark-provided' and 'mark = mask = 0'. IIUC before this fix, explicitly setting mark, mask to a) mark = 0, mask =
0xffffffffand b) mark = 0, mask = 0 behaved differently, whereas after this fix they will act the same.But as it seems there was never support for the (b) scenario and commit
9b42c1f179a6broke the existing behavior, so I'm ok with this fix.Thanks!
Eyal.Quoted 29 lines
Tested with additions to Android's kernel unit test suite:
https://android-review.googlesource.com/c/kernel/tests/+/860150Fixes 9b42c1f179a6 ("xfrm: Extend the output_mark to support inputdirection and masking") Signed-off-by: Benedict Wong <[email protected]> ---
net/xfrm/xfrm_policy.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)net/xfrm/xfrm_policy.c +4
-1@@ -2600,7 +2600,10 @@ static struct dst_entry *xfrm_bundle_create(struct xfrm_policy *policy, dst_copy_metrics(dst1, dst);if (xfrm[i]->props.mode !=XFRM_MODE_TRANSPORT ) { -__u32 mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]); +__u32 mark =0 ; + +if (xfrm[i]->props.smark.v || xfrm[i]->props.smark.m) + mark =xfrm_smark_get (fl->flowi_mark, xfrm[i]);family = xfrm[i]->props.family; dst = xfrm_dst_lookup(xfrm[i], tos,
fl->flowi_oif,
Re: [PATCH ipsec, resend 1/1] xfrm: Make set-mark default behavior backward compatible
In reply to Benedict WongQuoted 10 lines · Benedict Wong
Fixes
9b42c1f179a6, which changed the default route lookup behavior for tunnel mode SAs in the outbound direction to use the skb mark, whereas previously mark=0 was used if the output mark was unspecified. In mark-based routing schemes such as Android’s, this change in default behavior causes routing loops or lookup failures.This patch restores the default behavior of using a 0 mark while still incorporating the skb mark if the
SET_MARK(andSET_MARK_MASK) is specified.
Quoted 5 lines
Tested with additions to Android's kernel unit test suite:
https://android-review.googlesource.com/c/kernel/tests/+/860150Fixes 9b42c1f179a6 ("xfrm: Extend the output_mark to support input direction and masking")Signed-off-by Benedict Wong <[email protected]>
Applied, thanks a lot Benedict!