Skip to contentSkip to conversation

net/sched Bind logic fixes for cls_fw, cls_u32 and cls_route

PATCH In review valis 7 participants 19 messages Jul 21, 2023 – Jul 26, 2023

Subsystems (5)
lore
Share & tools
lore

valis

author
#0 lore raw

cover net/sched Bind logic fixes for cls_fw, cls_u32 and cls_route

Three classifiers (cls_fw, cls_u32 and cls_route) always copy tcf_result struct into the new instance of the filter on update.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

This patch set fixes this issue in all affected classifiers by no longer copying the tcf_result struct from the old filter.

valis (3):
net/sched: cls_u32: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_fw: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_route: No longer copy tcf_result on update to avoid
use-after-free

 net/sched/cls_fw.c    | 1 -
 net/sched/cls_route.c | 1 -
 net/sched/cls_u32.c   | 1 -
 3 files changed, 3 deletions(-)
Signature · 1 line
2.30.2
author
#1 lore raw

PATCH 1/3 net/sched: cls_u32: No longer copy tcf_result on update to avoid use-after-free

When u32_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
---
 net/sched/cls_u32.c | 1 -
 1 file changed, 1 deletion(-)
net/sched/cls_u32.c +0 -1
index 5abf31e432ca..19aa60d1eea7 100644
@@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
 
 	new->ifindex = n->ifindex;
 	new->fshift = n->fshift;
-	new->res = n->res;
 	new->flags = n->flags;
 	RCU_INIT_POINTER(new->ht_down, ht);
 
Signature · 1 line
2.30.2
#2 lore raw

Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on update to avoid use-after-free

Quoted 33 lines · valis

When u32_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
---
 net/sched/cls_u32.c | 1 -
 1 file changed, 1 deletion(-)
net/sched/cls_u32.c +0 -2
index 5abf31e432ca..19aa60d1eea7 100644
@@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
 
        new->ifindex = n->ifindex;
        new->fshift = n->fshift;
-       new->res = n->res;
        new->flags = n->flags;
        RCU_INIT_POINTER(new->ht_down, ht);
 
--
Quoted 2 lines

2.30.2

Hi,

We also thought it's also the correct fixes,
but we're not sure because it will always remove the already bound qdisc class when we change the filter, even tho we never specify the new TCA_U32_CLASSID in the new filter.

I also look at the implementation of cls_tcindex and cls_rsvp which still copy the old tcf_result, but don't call the tcf_unbind_filter when changing the filter.

If it's the intended behaviour, then I'm good with this patch.

Thanks & Regards,
M A Ramdhan

#3 lore raw

Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on update to avoid use-after-free

Quoted 38 lines · M A Ramdhan
Quoted 33 lines · valis

When u32_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
---
  net/sched/cls_u32.c | 1 -
  1 file changed, 1 deletion(-)
net/sched/cls_u32.c +0 -2
index 5abf31e432ca..19aa60d1eea7 100644
@@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
 
         new->ifindex = n->ifindex;
         new->fshift = n->fshift;
-       new->res = n->res;
         new->flags = n->flags;
         RCU_INIT_POINTER(new->ht_down, ht);
 
--
Quoted 2 lines

2.30.2

Hi,

Quoted 4 lines

We also thought it's also the correct fixes,
but we're not sure because it will always remove the already bound qdisc class when we change the filter, even tho we never specify the new TCA_U32_CLASSID in the new filter.

The user should always explicitly tell the classid. Some other classifiers are already behaving like this, these were just wrong.

Quoted 4 lines

I also look at the implementation of cls_tcindex and cls_rsvp which still copy the old tcf_result, but don't call the tcf_unbind_filter when changing the filter.

Both were deprecated and removed as they were beyond saving.

Quoted 5 lines

If it's the intended behaviour, then I'm good with this patch.

Thanks & Regards,
M A Ramdhan

#4 lore raw

Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on update to avoid use-after-free

Quoted 40 lines · Pedro Tammela
Quoted 38 lines · M A Ramdhan
Quoted 33 lines · valis

When u32_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
---
  net/sched/cls_u32.c | 1 -
  1 file changed, 1 deletion(-)
net/sched/cls_u32.c +0 -2
index 5abf31e432ca..19aa60d1eea7 100644
@@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
 
         new->ifindex = n->ifindex;
         new->fshift = n->fshift;
-       new->res = n->res;
         new->flags = n->flags;
         RCU_INIT_POINTER(new->ht_down, ht);
 
--
Quoted 2 lines

2.30.2

Hi,

Quoted 8 lines
Quoted 4 lines

We also thought it's also the correct fixes,
but we're not sure because it will always remove the already bound qdisc class when we change the filter, even tho we never specify the new TCA_U32_CLASSID in the new filter.

The user should always explicitly tell the classid. Some other classifiers are already behaving like this, these were just wrong.

Understand, thanks for the clarification.

Regards,
M A Ramdhan

Quoted 13 lines
Quoted 4 lines

I also look at the implementation of cls_tcindex and cls_rsvp which still copy the old tcf_result, but don't call the tcf_unbind_filter when changing the filter.

Both were deprecated and removed as they were beyond saving.

Quoted 5 lines

If it's the intended behaviour, then I'm good with this patch.

Thanks & Regards,
M A Ramdhan

maintainer
#5 lore raw

Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on update to avoid use-after-free

In reply to M A Ramdhan
Quoted 39 lines · M A Ramdhan
Quoted 33 lines · valis

When u32_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
---
 net/sched/cls_u32.c | 1 -
 1 file changed, 1 deletion(-)
net/sched/cls_u32.c +0 -2
index 5abf31e432ca..19aa60d1eea7 100644
@@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
 
        new->ifindex = n->ifindex;
        new->fshift = n->fshift;
-       new->res = n->res;
        new->flags = n->flags;
        RCU_INIT_POINTER(new->ht_down, ht);
 
--
Quoted 2 lines

2.30.2

Hi,

Quoted 4 lines

We also thought it's also the correct fixes,
but we're not sure because it will always remove the already bound qdisc class when we change the filter, even tho we never specify the new TCA_U32_CLASSID in the new filter.

I am assuming you are referring to the u32 classifier here and from what you are describing you first create a filter with reference to an existing class and then you replace it to not reference the class - am i correct? If yes, then please provide an example policy setup? We tested this scenario on u32 extensively and it should be fine.

cheers,
jamal

Quoted 8 lines

I also look at the implementation of cls_tcindex and cls_rsvp which still copy the old tcf_result, but don't call the tcf_unbind_filter when changing the filter.

If it's the intended behaviour, then I'm good with this patch.

Thanks & Regards,
M A Ramdhan

author
#6 lore raw

PATCH 2/3 net/sched: cls_fw: No longer copy tcf_result on update to avoid use-after-free

In reply to valis

When fw_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes e35a8ee5993b ("net: sched: fw use RCU")
---
 net/sched/cls_fw.c | 1 -
 1 file changed, 1 deletion(-)
net/sched/cls_fw.c +0 -1
index 8641f8059317..c49d6af0e048 100644
@@ -267,7 +267,6 @@ static int fw_change(struct net *net, struct sk_buff *in_skb,
 			return -ENOBUFS;
 
 		fnew->id = f->id;
-		fnew->res = f->res;
 		fnew->ifindex = f->ifindex;
 		fnew->tp = f->tp;
 
Signature · 1 line
2.30.2
author
#7 lore raw

PATCH 3/3 net/sched: cls_route: No longer copy tcf_result on update to avoid use-after-free

In reply to valis

When route4_change() is called on an existing filter, the whole tcf_result struct is always copied into the new instance of the filter.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

Fix this by no longer copying the tcf_result struct from the old filter.

Fixes 1109c00547fc ("net: sched: RCU cls_route")
---
 net/sched/cls_route.c | 1 -
 1 file changed, 1 deletion(-)
net/sched/cls_route.c +0 -1
index d0c53724d3e8..1e20bbd687f1 100644
@@ -513,7 +513,6 @@ static int route4_change(struct net *net, struct sk_buff *in_skb,
 	if (fold) {
 		f->id = fold->id;
 		f->iif = fold->iif;
-		f->res = fold->res;
 		f->handle = fold->handle;
 
 		f->tp = fold->tp;
Signature · 1 line
2.30.2
Reviewed-by Tested-by on the v1 series
#8 lore raw
In reply to valis
Quoted 19 lines · valis

Three classifiers (cls_fw, cls_u32 and cls_route) always copy tcf_result struct into the new instance of the filter on update.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

This patch set fixes this issue in all affected classifiers by no longer copying the tcf_result struct from the old filter.

valis (3):
net/sched: cls_u32: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_fw: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_route: No longer copy tcf_result on update to avoid
use-after-free

Quoted 5 lines
  net/sched/cls_fw.c    | 1 -
  net/sched/cls_route.c | 1 -
  net/sched/cls_u32.c   | 1 -
  3 files changed, 3 deletions(-)

For the series,

maintainer Acked-by
#9 lore raw
Quoted 27 lines · Pedro Tammela
Quoted 19 lines · valis

Three classifiers (cls_fw, cls_u32 and cls_route) always copy tcf_result struct into the new instance of the filter on update.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

This patch set fixes this issue in all affected classifiers by no longer copying the tcf_result struct from the old filter.

valis (3):
net/sched: cls_u32: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_fw: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_route: No longer copy tcf_result on update to avoid
use-after-free

Quoted 5 lines
  net/sched/cls_fw.c    | 1 -
  net/sched/cls_route.c | 1 -
  net/sched/cls_u32.c   | 1 -
  3 files changed, 3 deletions(-)
Quoted 4 lines

For the series,

For the series:

cheers,
jamal

Reviewed-by Tested-by
#10 lore raw
Quoted 29 lines · Jamal Hadi Salim
Quoted 27 lines · Pedro Tammela
Quoted 24 lines · valis

Three classifiers (cls_fw, cls_u32 and cls_route) always copy tcf_result struct into the new instance of the filter on update.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

This patch set fixes this issue in all affected classifiers by no longer copying the tcf_result struct from the old filter.

valis (3):
net/sched: cls_u32: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_fw: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_route: No longer copy tcf_result on update to avoid
use-after-free

  net/sched/cls_fw.c    | 1 -
  net/sched/cls_route.c | 1 -
  net/sched/cls_u32.c   | 1 -
  3 files changed, 3 deletions(-)
Quoted 7 lines
Quoted 4 lines

For the series,

For the series:

For the series:

Regards,
M A Ramdhan

Quoted 3 lines

cheers,
jamal

maintainer
#11 lore raw
In reply to valis

Hi,

Quoted 11 lines · valis

Three classifiers (cls_fw, cls_u32 and cls_route) always copy tcf_result struct into the new instance of the filter on update.

This causes a problem when updating a filter bound to a class, as tcf_unbind_filter() is always called on the old instance in the success path, decreasing filter_cnt of the still referenced class and allowing it to be deleted, leading to a use-after-free.

This patch set fixes this issue in all affected classifiers by no longer copying the tcf_result struct from the old filter.

Quoted 7 lines

valis (3):
net/sched: cls_u32: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_fw: No longer copy tcf_result on update to avoid
use-after-free
net/sched: cls_route: No longer copy tcf_result on update to avoid
use-after-free

The SoB in used here sounds really like a pseudonym, which in turn is not explicitly forbidden by the the process, but a is IMHO a bit borderline:

https://elixir.bootlin.com/linux/v6.4.5/source/Documentation/process/submitting-patches.rst#L415

@valis: could you please re-submit this using your a more identificative account? You can retain the already collected acks.

Thanks!

Paolo

Quoted 6 lines
 net/sched/cls_fw.c    | 1 -
 net/sched/cls_route.c | 1 -
 net/sched/cls_u32.c   | 1 -
 3 files changed, 3 deletions(-)
author Acked-by Greg KH Acked-by Michael Dolan
#12 lore raw
Quoted 11 lines · Paolo Abeni

Hi,

The SoB in used here sounds really like a pseudonym, which in turn is not explicitly forbidden by the the process, but a is IMHO a bit borderline:

https://elixir.bootlin.com/linux/v6.4.5/source/Documentation/process/submitting-patches.rst#L415

@valis: could you please re-submit this using your a more identificative account? You can retain the already collected acks.

Hi Paolo!

The document you quoted does not forbid pseudonyms. In fact, it was recently updated to clarify that very fact.

You might want to take a look at this commit:

commit d4563201f33a022fc0353033d9dfeb1606a88330

Author: Linus Torvalds <[email protected]>
Date:   Sun Feb 26 11:25:04 2023 -0800

Documentation: simplify and clarify DCO contribution example language

Long long ago, in a more innocent time, Greg wrote the clarification for how the DCO should work and that you couldn't make anonymous contributions, because the sign-off needed to be something we could check back with.

It was 2006, and nobody reacted to the wording, the whole Facebook 'real name' controversy was a decade in the future, and nobody even thought about it. And despite the language, we've always accepted nicknames and that language was never meant to be any kind of exclusionary wording.

In fact, even when it became a discussion in other adjacent projects, apparently nobody even thought to just clarify the language in the kernel docs, and instead we had projects like the CNCF that had long discussions about it, and wrote their own clarifications [1] of it.

Just simplify the wording to the point where it shouldn't be causing unnecessary angst and pain, or scare away people who go by preferred naming.

[1]
Fixes af45f32d25cc ("We can not allow anonymous contributions to

the kernel")

(quoted diff) +0 -0

b/Documentation/process/submitting-patches.rst
index 7dc94555417d..fab44ae732e3 100644

Documentation/process/submitting-patches.rst +1 -1
@@ -407,7 +407,7 @@ then you just add a line saying::
 
        Signed-off-by: Random J Developer <[email protected]>
 
-using your real name (sorry, no pseudonyms or anonymous contributions.)
+using a known identity (sorry, no anonymous contributions.)
 
 

Best regards,

valis

maintainer
#13 lore raw
Quoted 2 lines · On Tue, 25 Jul 2023 21:05:23 +0200 valis wrote:

The document you quoted does not forbid pseudonyms. In fact, it was recently updated to clarify that very fact.

We don't know who you are. To my understanding the adjustment means that you are not obligated to use the name on your birth certificate but we need to know who you are.

Why is it always "security" people who try act like this is some make believe metaverse. We're working on a real project with real licenses and real legal implications.

Your S-o-b is pretty much meaningless. If a "real" person can vouch for who you are or put their own S-o-b on your code that's fine.

author
#14 lore raw
Quoted 4 lines · Jakub Kicinski
Quoted 2 lines · On Tue, 25 Jul 2023 21:05:23 +0200 valis wrote:

The document you quoted does not forbid pseudonyms. In fact, it was recently updated to clarify that very fact.

Hi Jakub!

Quoted 4 lines

We don't know who you are. To my understanding the adjustment means that you are not obligated to use the name on your birth certificate but we need to know who you are.

I could start a discussion about what makes a name valid, but I'm pretty sure netdev is not the right place for it.

Quoted 7 lines

Why is it always "security" people who try act like this is some make believe metaverse. We're working on a real project with real licenses and real legal implications.

Your S-o-b is pretty much meaningless. If a "real" person can vouch for who you are or put their own S-o-b on your code that's fine.

I posted my patches to this mailing list per maintainer's request and according to the published rules of the patch submission process as I understood them.
Sorry if I misinterpreted something and wasted anybody's time.

I'm not going to resubmit the patches under a different name.

Please feel free to use them if someone is willing to sign off on them.

Best regards,

valis

maintainer
#15 lore raw
Quoted 7 lines · On Tue, 25 Jul 2023 23:36:14 +0200 valis wrote:
Quoted 3 lines · Jakub Kicinski

We don't know who you are. To my understanding the adjustment means that you are not obligated to use the name on your birth certificate but we need to know who you are.

I could start a discussion about what makes a name valid, but I'm pretty sure netdev is not the right place for it.

Agreed, I CCed Greg KH to keep me honest, in case I'm outright incorrect. If it's a gray zone kinda answer I'm guessing that nobody here really wants to spend time discussing it.

Quoted 7 lines
Quoted 6 lines

Why is it always "security" people who try act like this is some make believe metaverse. We're working on a real project with real licenses and real legal implications.

Your S-o-b is pretty much meaningless. If a "real" person can vouch for who you are or put their own S-o-b on your code that's fine.

Quoted 8 lines

I posted my patches to this mailing list per maintainer's request and according to the published rules of the patch submission process as I understood them.
Sorry if I misinterpreted something and wasted anybody's time.

I'm not going to resubmit the patches under a different name.

Please feel free to use them if someone is willing to sign off on them.

If Jamal or anyone else feels like they can vouch for the provenance of the code, I think that may be the best compromise.

#16 lore raw
Quoted 6 lines · Jakub Kicinski
Quoted 5 lines · On Tue, 25 Jul 2023 23:36:14 +0200 valis wrote:
Quoted 3 lines · Jakub Kicinski

We don't know who you are. To my understanding the adjustment means that you are not obligated to use the name on your birth certificate but we need to know who you are.

Quoted 6 lines
Quoted 2 lines

I could start a discussion about what makes a name valid, but I'm pretty sure netdev is not the right place for it.

Agreed, I CCed Greg KH to keep me honest, in case I'm outright incorrect. If it's a gray zone kinda answer I'm guessing that nobody here really wants to spend time discussing it.

You are not incorrect.

Sorry "valis", we need more information to be able to take patches from you as an author for obvious reasons.

thanks,

greg k-h

maintainer
#17 lore raw
In reply to Jakub Kicinski
Quoted 26 lines · Jakub Kicinski
Quoted 7 lines · On Tue, 25 Jul 2023 23:36:14 +0200 valis wrote:
Quoted 3 lines · Jakub Kicinski

We don't know who you are. To my understanding the adjustment means that you are not obligated to use the name on your birth certificate but we need to know who you are.

I could start a discussion about what makes a name valid, but I'm pretty sure netdev is not the right place for it.

Agreed, I CCed Greg KH to keep me honest, in case I'm outright incorrect. If it's a gray zone kinda answer I'm guessing that nobody here really wants to spend time discussing it.

Quoted 12 lines
Quoted 6 lines

Why is it always "security" people who try act like this is some make believe metaverse. We're working on a real project with real licenses and real legal implications.

Your S-o-b is pretty much meaningless. If a "real" person can vouch for who you are or put their own S-o-b on your code that's fine.

I posted my patches to this mailing list per maintainer's request and according to the published rules of the patch submission process as I understood them.
Sorry if I misinterpreted something and wasted anybody's time.

Quoted 6 lines
Quoted 3 lines

I'm not going to resubmit the patches under a different name.

Please feel free to use them if someone is willing to sign off on them.

If Jamal or anyone else feels like they can vouch for the provenance of the code, I think that may be the best compromise.

I am more than happy to add my SoB - these are real issues being fixed and this is not the first time that Valis identified and helped resolve legit issues in tc.

cheers,
jamal

maintainer
#18 lore raw
Quoted 6 lines · On Wed, 26 Jul 2023 09:59:02 -0400 Jamal Hadi Salim wrote:
Quoted 2 lines

If Jamal or anyone else feels like they can vouch for the provenance of the code, I think that may be the best compromise.

I am more than happy to add my SoB - these are real issues being fixed and this is not the first time that Valis identified and helped resolve legit issues in tc.

Great! Please repost with your signoffs.