bpf: Relax 8 frame limitation for global subprogs
Subsystems (4)
Emil Tsalapatis
reviewerv3 cover bpf: Relax 8 frame limitation for global subprogs
The BPF verifier currently limits the maximum runtime call stack to 8 frames. Larger BPF programs like sched-ext schedulers routinely fail verification because they exceed this limit, even as they use very little actual stack space for each frame.
Relax the verifier to permit call stacks > 8 frames deep when the call stacks include global subprogs. The old 8 stack frame limit now only applies to call stacks composed entirely of static function calls. This works because global functions are each verified in isolation, so the verifier does not need to cross-reference verification state across the function call boundary, which has been the reason for limiting the call stack size in the first place.
This patch does not change the verification time limit of 8 stack frames. Static functions that are inlined for verification purposes still only go 8 frames deep to avoid changing the verifier's internal data structures used for verification. These data structures only support holding information on up to 8 stack frames.
This patch also does not adjust the actual maximum stack size of 512.
CHANGELOG
=========
v2 -> v3 (https://lore.kernel.org/bpf/[email protected]/)
- Change logic to remove arbitrary limit on call depth (Eduard)
- Add additional selftests (Eduard)
v1 -> v2 (https://lore.kernel.org/bpf/[email protected])
- Adjust patch to only increase the runtime stack depth, leaving the verification-time stack depth unchanged (Alexei)
Emil Tsalapatis (2):
bpf: Only enforce 8 frame call stack limit for all-static stacks
bpf: Add deep call stack selftests
include/linux/bpf_verifier.h | 6 ++ kernel/bpf/verifier.c | 43 ++++++---- .../bpf/prog_tests/test_global_funcs.c | 2 + .../selftests/bpf/progs/test_global_func3.c | 18 ++--- .../bpf/progs/test_global_func_deep_stack.c | 79 +++++++++++++++++++ 5 files changed, 123 insertions(+), 25 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/test_global_func_deep_stack.c
Signature · 1 line
2.49.0
Emil Tsalapatis
reviewerPATCH v3 1/2 bpf: Only enforce 8 frame call stack limit for all-static stacks
The BPF verifier currently enforces a call stack depth of 8 frames, regardless of the actual stack space consumption of those frames. The limit is necessary for static call stacks, because the bookkeeping data structures used by the verifier when stepping into static functions during verification only support 8 stack frames. However, this limitation only matters for static stack frames: Global subprogs are verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification now only fails when there is a sequence of 8 calls to non-global subprogs. Calling into a global subprog resets the counter. This allows deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only the maximum number of frames we can place in it.
Also change the progs/test_global_func3.c selftest to use static functions, since with the new patch it would otherwise unexpectedly pass verification.
--- include/linux/bpf_verifier.h | 6 +++ kernel/bpf/verifier.c | 43 ++++++++++++------- .../selftests/bpf/progs/test_global_func3.c | 18 ++++---- 3 files changed, 42 insertions(+), 25 deletions(-)
include/linux/bpf_verifier.h +6 -0
@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {/* 'start' has to be the first field otherwise find_subprog() won't work */ u32 start;/* insn idx of function entry point */
@@ -677,6 +679,10 @@ struct bpf_subprog_info {enum priv_stack_mode priv_stack_mode;struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS ]; + +int ret_insn; +int frame; +int cidx; };struct bpf_verifier_env ;
kernel/bpf/verifier.c +27 -16
98 lines not on this page · raw message
tools/testing/selftests/bpf/progs/test_global_func3.c +9 -9
@@ -5,56 +5,56 @@#include <bpf/bpf_helpers.h> #include "bpf_misc.h" -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f1 (struct __sk_buff *skb) {return skb->len; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f2 (int val,struct __sk_buff *skb) {return f1 (skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f3 (int val,struct __sk_buff *skb,int var) {return f2 (var, skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f4 (struct __sk_buff *skb) {return f3 (1 , skb,2 ); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f5 (struct __sk_buff *skb) {return f4 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f6 (struct __sk_buff *skb) {return f5 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f7 (struct __sk_buff *skb) {return f6 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f8 (struct __sk_buff *skb) {return f7 (skb); } SEC("tc") -__failure __msg("the call stack of 8 frames") +__failure __msg("the call stack of 9 frames") int global_func3(struct __sk_buff *skb) {return f8 (skb);
Signature · 1 line
2.49.0
[email protected]
AI reviewRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
This isn't a bug, but the forward declaration above the struct definition appears unnecessary. The new fields (ret_insn, frame, cidx) are all plain int types, not self-referential pointers that…
include/linux/bpf_verifier.h enum priv_stack_mode { · line 655
Quoted 10 lines
include/linux/bpf_verifier.h +2
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {
This isn't a bug, but the forward declaration above the struct definition appears unnecessary. The new fields (ret_insn, frame, cidx) are all plain int types, not self-referential pointers that would require forward declaration. Is there a reason for it?
Quoted 3 lines
kernel/bpf/verifier.c +0
-0
[ ... ]
kernel/bpf/verifier.c line 6814
Quoted 9 lines
kernel/bpf/verifier.c +1
-1@@ -6799,7 +6808,7 @@if (subprog[idx].has_tail_call) tail_call_reachable = true; - frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);
In the selftest below, the expected error message was changed to "the call stack of 9 frames", but this verbose() call still prints the raw frame counter.
With 9 static functions in the chain (global_func3 + f1 through f8), frame reaches 8 when entering f1, and the verifier outputs "the call stack of 8 frames is too deep !" -- which does not contain "9 frames".
The MAX_BPF_STACK error path was updated from frame + 1 to a cidx chain walk for total count:
for (total = 1; subprog[idx].cidx >= 0 ; total++)
idx = subprog[idx].cidx;
verbose(env, "combined stack size of %d calls ...\n", total, depth);Was a similar update intended here? Changing frame to frame + 1 in the verbose() call would match the selftest expectation and be consistent with how the stack overflow message now reports total count.
Quoted 3 lines
tools/testing/selftests/bpf/progs/test_global_func3.c +0
-0
[ ... ]
tools/testing/selftests/bpf/progs/test_global_func3.c line 9
Quoted 7 lines
tools/testing/selftests/bpf/progs/test_global_func3.c +1
-1@@ -5,56 +5,56 @@ SEC("tc") -__failure __msg("the call stack of 8 frames") +__failure __msg("the call stack of 9 frames")int global_func3 (struct __sk_buff *skb) {return f8 (skb);
This expected message does not match the verifier output as described above.
Also, the function header comment for check_max_stack_depth_subprog() currently reads:
Since recursion is prevented by check_cfg() this algorithm only needs a local stack of MAX_CALL_FRAMES to remember callsites
The local arrays ret_insn[] and ret_prog[] have been removed in favor of struct fields (subprog[].cidx, subprog[].frame, subprog[].ret_insn). Should this comment be updated to reflect the new approach? The v2 patch did update this comment for its design.
---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md
CI run summary: https://github.com/kernel-patches/bpf/actions/runs/22608683847
Emil Tsalapatis
reviewerRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
Quoted 7 lines · bot+bpf-ci
Quoted 7 lines
include/linux/bpf_verifier.h +0
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, };
Quoted 9 lines
include/linux/bpf_verifier.h enum priv_stack_mode { · line 655
Quoted 3 lines
include/linux/bpf_verifier.h +2
-0+struct bpf_subprog_info ; +struct bpf_subprog_info {This isn't a bug, but the forward declaration above the struct definition appears unnecessary. The new fields (
ret_insn, frame, cidx) are all plain int types, not self-referential pointers that would require forward declaration. Is there a reason for it?
Ack, it _used_ to be self referential and I forgot to remove it. Will fix.
Quoted 27 lines
Quoted 3 lines
kernel/bpf/verifier.c +0
-0[ ... ]
kernel/bpf/verifier.c line 6814
Quoted 9 lines
kernel/bpf/verifier.c +1
-1@@ -6799,7 +6808,7 @@if (subprog[idx].has_tail_call) tail_call_reachable = true; - frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);In the selftest below, the expected error message was changed to "the call stack of 9 frames", but this
verbose()call still prints the raw frame counter.With 9 static functions in the chain (
global_func3+ f1 through f8), frame reaches 8 when entering f1, and the verifier outputs "the call stack of 8 frames is too deep !" -- which does not contain "9 frames".The
MAX_BPF_STACKerror path was updated from frame + 1 to a cidx chain walk for total count:
Quoted 8 lines
for (total = 1; subprog[idx].cidx >= 0 ; total++) idx = subprog[idx].cidx; verbose(env, "combined stack size of %d calls ...\n", total, depth);Was a similar update intended here? Changing frame to frame + 1 in the
verbose()call would match the selftest expectation and be consistent with how the stack overflow message now reports total count.
No, that one should still be either 8 or 9 (see below) because it's the actual limit - including the global frames in the tally would be confusing. Maybe the right thing is to adjust the commit message.
Quoted 20 lines
Quoted 3 lines
tools/testing/selftests/bpf/progs/test_global_func3.c +0
-0[ ... ]
tools/testing/selftests/bpf/progs/test_global_func3.c line 9
Quoted 7 lines
tools/testing/selftests/bpf/progs/test_global_func3.c +1
-1@@ -5,56 +5,56 @@ SEC("tc") -__failure __msg("the call stack of 8 frames") +__failure __msg("the call stack of 9 frames")int global_func3 (struct __sk_buff *skb) {return f8 (skb);This expected message does not match the verifier output as described above.
Also, the function header comment for
check_max_stack_depth_subprog()currently reads:
Quoted 8 lines
Since recursion is prevented by
check_cfg()this algorithm only needs a local stack ofMAX_CALL_FRAMESto remember callsitesThe local arrays ret_insn[] and ret_prog[] have been removed in favor of struct fields (subprog[].cidx, subprog[].frame, subprog[].ret_insn). Should this comment be updated to reflect the new approach? The v2 patch did update this comment for its design.
It does though, all tests are passing: There are two almost identical messages in the verifier that report off-by-one different numbers:
"the call stack of %d frames is too deep", (setup_func_entry) "the call stack of %d frames is too deep !" (check_max_stack_depth_subprog)
The one triggering now is in setup_func_entry inside do_check_{main, subprogs}, so the number of frames reported has changed. Expecting 9 frames is consistent with other selftests.
Quoted 6 lines
---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.mdCI run summary: https://github.com/kernel-patches/bpf/actions/runs/22608683847
Eduard Zingerman
maintainerRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
In reply to Emil TsalapatisQuoted 41 lines · Emil Tsalapatis
The BPF verifier currently enforces a call stack depth of 8 frames, regardless of the actual stack space consumption of those frames. The limit is necessary for static call stacks, because the bookkeeping data structures used by the verifier when stepping into static functions during verification only support 8 stack frames. However, this limitation only matters for static stack frames: Global subprogs are verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification now only fails when there is a sequence of 8 calls to non-global subprogs. Calling into a global subprog resets the counter. This allows deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only the maximum number of frames we can place in it.
Also change the
progs/test_global_func3.cselftest to use static functions, since with the new patch it would otherwise unexpectedly pass verification.Signed-off-by Emil Tsalapatis <[email protected]>--- include/linux/bpf_verifier.h | 6 +++ kernel/bpf/verifier.c | 43 ++++++++++++------- .../selftests/bpf/progs/test_global_func3.c | 18 ++++---- 3 files changed, 42 insertions(+), 25 deletions(-)include/linux/bpf_verifier.h +2
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {/* 'start' has to be the first field otherwise find_subprog() won't work */ u32 start;/* insn idx of function entry point */
include/linux/bpf_verifier.h struct bpf_subprog_info { · line 682
Quoted 4 lines
include/linux/bpf_verifier.h +1
-0@@ -677,6 +679,10 @@ struct bpf_subprog_info {enum priv_stack_mode priv_stack_mode;struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS ]; +
Nit: please add a comment here saying the below are temporary values
used in check_max_stack_depth_subprog() (tbh, I'd move those to a separate array in env but maybe that's an overkill).
include/linux/bpf_verifier.h struct bpf_subprog_info { · line 685
Quoted 3 lines
include/linux/bpf_verifier.h +3
-0+int ret_insn; +int frame; +int cidx;
Nit: please rename to caller and add a comment.
Quoted 36 lines
include/linux/bpf_verifier.h +0
-0};struct bpf_verifier_env ;kernel/bpf/verifier.c +9
-4@@ -6652,9 +6652,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,struct bpf_insn *insn = env->prog->insnsi;int depth =0 , frame =0 , i, subprog_end, subprog_depth;bool tail_call_reachable = false; -int ret_insn[MAX_CALL_FRAMES ]; -int ret_prog[MAX_CALL_FRAMES ]; -int j; +int total; +int tmp; + +/* no caller idx */ + subprog[idx].cidx =-1 ; i = subprog[idx].start;if (!priv_stack_supported)@@ -6706,8 +6708,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, }else { depth += subprog_depth;if (depth >MAX_BPF_STACK ) { +for (total =1 ; subprog[idx].cidx >=0 ; total++) + idx = subprog[idx].cidx; +verbose (env,"combined stack size of %d calls is %d. Too large\n" , - frame +1 , depth); + total, depth);return -EACCES ; } }
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6730
Quoted 3 lines
kernel/bpf/verifier.c +0
-0@@ -6723,8 +6728,8 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,continue ;if (subprog[idx].is_cb) err = true;
Below the old version of the loop does not include current frame, but the new version *does* include current frame. Looks like the above check can be removed.
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6735
Quoted 7 lines
kernel/bpf/verifier.c +2
-2-for (int c =0 ; c < frame && !err; c++) { -if (subprog[ret_prog[c]].is_cb) { +for (tmp = idx; tmp >=0 && !err; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_cb) { err = true;break ; }
This code checks that bpf_throw() cannot be called from callbacks. The is_cb is set in push_callback_call() both for sync and async callbacks, it is also set for exception callback separately. Meaning that check_return_code() can be simplified further.
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6750
Quoted 9 lines
kernel/bpf/verifier.c +3
-2@@ -6740,8 +6745,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (!bpf_pseudo_call (insn + i) && !bpf_pseudo_func (insn + i))continue ;/* remember insn and function to return to */ - ret_insn[frame] = i +1 ; - ret_prog[frame] = idx; + + subprog[idx].frame = frame; + subprog[idx].ret_insn = i +1 ;
Nit: move this down to subprog[sidx].cidx = idx;, so that the whole
"frame" setup is done in one place?
Quoted 23 lines
kernel/bpf/verifier.c +4
-1/* find the callee */ next_insn = i + insn[i].imm +1 ;@@ -6762,6 +6768,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, } } i = next_insn; + +/* caller idx */ + subprog[sidx].cidx = idx; idx = sidx;if (!priv_stack_supported) subprog[idx].priv_stack_mode =NO_PRIV_STACK ;@@ -6769,7 +6778,7 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (subprog[idx].has_tail_call) tail_call_reachable = true; - frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6794
Quoted 3 lines
kernel/bpf/verifier.c +0
-0@@ -6783,12 +6792,12 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,* tail call counter throughout bpf2bpf calls combined with tailcalls */ if (tail_call_reachable)
Unrelated to current patch, but looks like an issue. tail_call_reachable is not reset anywhere in check_max_stack_depth_subprog(). Once it flips to true all calls visited afterwards trigger spine to be marked.
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6796
Quoted 4 lines
kernel/bpf/verifier.c +2
-2-for (j =0 ; j < frame; j++) { -if (subprog[ret_prog[j]].is_exception_cb) { +for (tmp = idx; tmp >=0 ; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_exception_cb) {
As with previous such loop, the new code seem to include current frame. Does this change anything?
Quoted 28 lines
kernel/bpf/verifier.c +7
-5verbose (env,"cannot tail call within exception cb\n" );return -EINVAL ; } - subprog[ret_prog[j]].tail_call_reachable = true; + subprog[tmp].tail_call_reachable = true; }if (subprog[0 ].tail_call_reachable) env->prog->aux->tail_call_reachable = true;@@ -6796,13 +6805,15 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,/* end of for() loop means the last insn of the 'subprog' * was reached. Doesn't matter whether it was JA or EXIT */ -if (frame ==0 ) +if (frame ==0 && subprog[idx].cidx <0 )return 0 ;if (subprog[idx].priv_stack_mode !=PRIV_STACK_ADAPTIVE ) depth -=round_up_stack_depth (env, subprog[idx].stack_depth); - frame--; - i = ret_insn[frame]; - idx = ret_prog[frame]; + + idx = subprog[idx].cidx; + frame = subprog[idx].frame; + i = subprog[idx].ret_insn; + goto continue_func; }
[...]
Emil Tsalapatis
reviewerRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
Quoted 51 lines · Eduard Zingerman
Quoted 41 lines · Emil Tsalapatis
The BPF verifier currently enforces a call stack depth of 8 frames, regardless of the actual stack space consumption of those frames. The limit is necessary for static call stacks, because the bookkeeping data structures used by the verifier when stepping into static functions during verification only support 8 stack frames. However, this limitation only matters for static stack frames: Global subprogs are verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification now only fails when there is a sequence of 8 calls to non-global subprogs. Calling into a global subprog resets the counter. This allows deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only the maximum number of frames we can place in it.
Also change the
progs/test_global_func3.cselftest to use static functions, since with the new patch it would otherwise unexpectedly pass verification.Signed-off-by Emil Tsalapatis <[email protected]>--- include/linux/bpf_verifier.h | 6 +++ kernel/bpf/verifier.c | 43 ++++++++++++------- .../selftests/bpf/progs/test_global_func3.c | 18 ++++---- 3 files changed, 42 insertions(+), 25 deletions(-)include/linux/bpf_verifier.h +2
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {/* 'start' has to be the first field otherwise find_subprog() won't work */ u32 start;/* insn idx of function entry point */ include/linux/bpf_verifier.h struct bpf_subprog_info { · line 682
Quoted 4 lines
include/linux/bpf_verifier.h +1
-0@@ -677,6 +679,10 @@ struct bpf_subprog_info {enum priv_stack_mode priv_stack_mode;struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS ]; +Nit: please add a comment here saying the below are temporary values
used incheck_max_stack_depth_subprog()(tbh, I'd move those to a separate array in env but maybe that's an overkill).
Quoted 7 lines
include/linux/bpf_verifier.h struct bpf_subprog_info { · line 685
Quoted 3 lines
include/linux/bpf_verifier.h +3
-0+int ret_insn; +int frame; +int cidx;Nit: please rename to
callerand add a comment.Quoted 1 line
include/linux/bpf_verifier.h +0
-0};
Ack (and to all other nits).
Quoted 34 lines
Quoted 34 lines
include/linux/bpf_verifier.h +0
-0struct bpf_verifier_env ;kernel/bpf/verifier.c +9
-4@@ -6652,9 +6652,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,struct bpf_insn *insn = env->prog->insnsi;int depth =0 , frame =0 , i, subprog_end, subprog_depth;bool tail_call_reachable = false; -int ret_insn[MAX_CALL_FRAMES ]; -int ret_prog[MAX_CALL_FRAMES ]; -int j; +int total; +int tmp; + +/* no caller idx */ + subprog[idx].cidx =-1 ; i = subprog[idx].start;if (!priv_stack_supported)@@ -6706,8 +6708,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, }else { depth += subprog_depth;if (depth >MAX_BPF_STACK ) { +for (total =1 ; subprog[idx].cidx >=0 ; total++) + idx = subprog[idx].cidx; +verbose (env,"combined stack size of %d calls is %d. Too large\n" , - frame +1 , depth); + total, depth);return -EACCES ; } }
Quoted 9 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6730
Quoted 4 lines
kernel/bpf/verifier.c +0
-0@@ -6723,8 +6728,8 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,continue ;if (subprog[idx].is_cb) err = true;Below the old version of the loop does not include current frame, but the new version *does* include current frame. Looks like the above check can be removed.
Makes sense.
Quoted 4 lines
Quoted 4 lines
kernel/bpf/verifier.c +2
-2-for (int c =0 ; c < frame && !err; c++) { -if (subprog[ret_prog[c]].is_cb) { +for (tmp = idx; tmp >=0 && !err; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_cb) {
Quoted 9 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6735
Quoted 3 lines
kernel/bpf/verifier.c +0
-0err = true;break ; }This code checks that
bpf_throw()cannot be called from callbacks. Theis_cbis set inpush_callback_call()both for sync and async callbacks, it is also set for exception callback separately. Meaning thatcheck_return_code()can be simplified further.
Do you mean we can simplify the check for choosing between check_return_code and check_global_subprog_return_code to just check is_cb instead of checking both is_async_cb and is_exception_cb?
Quoted 30 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6750
Quoted 9 lines
kernel/bpf/verifier.c +3
-2@@ -6740,8 +6745,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (!bpf_pseudo_call (insn + i) && !bpf_pseudo_func (insn + i))continue ;/* remember insn and function to return to */ - ret_insn[frame] = i +1 ; - ret_prog[frame] = idx; + + subprog[idx].frame = frame; + subprog[idx].ret_insn = i +1 ;Nit: move this down to
subprog[sidx].cidx = idx;, so that the whole
"frame" setup is done in one place?Quoted 17 lines
kernel/bpf/verifier.c +3
-0/* find the callee */ next_insn = i + insn[i].imm +1 ;@@ -6762,6 +6768,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, } } i = next_insn; + +/* caller idx */ + subprog[sidx].cidx = idx; idx = sidx;if (!priv_stack_supported) subprog[idx].priv_stack_mode =NO_PRIV_STACK ;@@ -6769,7 +6778,7 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (subprog[idx].has_tail_call) tail_call_reachable = true;
Quoted 15 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6794
Quoted 9 lines
kernel/bpf/verifier.c +1
-1- frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);@@ -6783,12 +6792,12 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,* tail call counter throughout bpf2bpf calls combined with tailcalls */ if (tail_call_reachable)Unrelated to current patch, but looks like an issue.
tail_call_reachableis not reset anywhere incheck_max_stack_depth_subprog(). Once it flips to true all calls visited afterwards trigger spine to be marked.
You're right, this is never unset. I can send a followup patchset for this along with the followups for the check_return_code patch. I believe we can change this to eagerly set the subprog's tail_call_reachable flag directly instead of using a stack variable.
Quoted 8 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6796
Quoted 4 lines
kernel/bpf/verifier.c +2
-2-for (j =0 ; j < frame; j++) { -if (subprog[ret_prog[j]].is_exception_cb) { +for (tmp = idx; tmp >=0 ; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_exception_cb) {As with previous such loop, the new code seem to include current frame. Does this change anything?
I think it is ok, though since as you pointed out tail_call_reachable doesn't currently seem to work correctly I can only assume: We set tail_call_reachable if any of the subprogs called by the current one make a tail call. If that's the case, the current frame should also count as tail_call_reachable.
It all depends on where tail_call_reachable is supposed to be set and unset. I think it's supposed to flow from callee to caller, and if that's the case we should hoist up the whole loop to where we initially set the stack variable.
Quoted 30 lines
Quoted 28 lines
kernel/bpf/verifier.c +7
-5verbose (env,"cannot tail call within exception cb\n" );return -EINVAL ; } - subprog[ret_prog[j]].tail_call_reachable = true; + subprog[tmp].tail_call_reachable = true; }if (subprog[0 ].tail_call_reachable) env->prog->aux->tail_call_reachable = true;@@ -6796,13 +6805,15 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,/* end of for() loop means the last insn of the 'subprog' * was reached. Doesn't matter whether it was JA or EXIT */ -if (frame ==0 ) +if (frame ==0 && subprog[idx].cidx <0 )return 0 ;if (subprog[idx].priv_stack_mode !=PRIV_STACK_ADAPTIVE ) depth -=round_up_stack_depth (env, subprog[idx].stack_depth); - frame--; - i = ret_insn[frame]; - idx = ret_prog[frame]; + + idx = subprog[idx].cidx; + frame = subprog[idx].frame; + i = subprog[idx].ret_insn; + goto continue_func; }[...]
Eduard Zingerman
maintainerRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
On Thu, 2026-03-05 at 14:38 -0500, Emil Tsalapatis wrote:
[...]
Quoted 26 lines
Quoted 9 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6730
Quoted 4 lines
kernel/bpf/verifier.c +0
-0@@ -6723,8 +6728,8 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,continue ;if (subprog[idx].is_cb) err = true;Below the old version of the loop does not include current frame, but the new version *does* include current frame. Looks like the above check can be removed.
Makes sense.
Quoted 4 lines
Quoted 4 lines
kernel/bpf/verifier.c +2
-2-for (int c =0 ; c < frame && !err; c++) { -if (subprog[ret_prog[c]].is_cb) { +for (tmp = idx; tmp >=0 && !err; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_cb) {Quoted 9 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6735
Quoted 3 lines
kernel/bpf/verifier.c +0
-0err = true;break ; }This code checks that
bpf_throw()cannot be called from callbacks. Theis_cbis set inpush_callback_call()both for sync and async callbacks, it is also set for exception callback separately. Meaning thatcheck_return_code()can be simplified further.
Quoted 3 lines
Do you mean we can simplify the check for choosing between
check_return_codeandcheck_global_subprog_return_codeto just checkis_cbinstead of checking bothis_async_cbandis_exception_cb?
Actually, I was thinking about frame->in_async_callback_fn checks in check_return_code(). But on a second thought these are unrelated. Please skip my comment above.
[...]
Quoted 30 lines
Quoted 10 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6794
Quoted 4 lines
kernel/bpf/verifier.c +0
-0@@ -6783,12 +6792,12 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,* tail call counter throughout bpf2bpf calls combined with tailcalls */ if (tail_call_reachable)Unrelated to current patch, but looks like an issue.
tail_call_reachableis not reset anywhere incheck_max_stack_depth_subprog(). Once it flips to true all calls visited afterwards trigger spine to be marked.You're right, this is never unset. I can send a followup patchset for this along with the followups for the
check_return_codepatch. I believe we can change this to eagerly set the subprog'stail_call_reachableflag directly instead of using a stack variable.Quoted 8 lines
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6796
Quoted 4 lines
kernel/bpf/verifier.c +2
-2-for (j =0 ; j < frame; j++) { -if (subprog[ret_prog[j]].is_exception_cb) { +for (tmp = idx; tmp >=0 ; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_exception_cb) {As with previous such loop, the new code seem to include current frame. Does this change anything?
I think it is ok, though since as you pointed out
tail_call_reachabledoesn't currently seem to work correctly I can only assume: We settail_call_reachableif any of the subprogs called by the current one make a tail call. If that's the case, the current frame should also count astail_call_reachable.
Quoted 4 lines
It all depends on where
tail_call_reachableis supposed to be set and unset. I think it's supposed to flow from callee to caller, and if that's the case we should hoist up the whole loop to where we initially set the stack variable.
So, the tail_call_reachable flag is transferred to subprog[*].tail_call_reachable property. In check_subprogs() there is direct subprog[cur_subprog].tail_call_reachable = true if tail call exists in the subprogram. Looks like behavior does not really change with this patch.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
kernel/bpf/verifier.c +1
-1verbose (env,"cannot tail call within exception cb\n" );return -EINVAL ; } - subprog[ret_prog[j]].tail_call_reachable = true; + subprog[tmp].tail_call_reachable = true; }if (subprog[0 ].tail_call_reachable) env->prog->aux->tail_call_reachable = true;
[...]
Re: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
In reply to Emil TsalapatisQuoted 41 lines · Emil Tsalapatis
The BPF verifier currently enforces a call stack depth of 8 frames, regardless of the actual stack space consumption of those frames. The limit is necessary for static call stacks, because the bookkeeping data structures used by the verifier when stepping into static functions during verification only support 8 stack frames. However, this limitation only matters for static stack frames: Global subprogs are verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification now only fails when there is a sequence of 8 calls to non-global subprogs. Calling into a global subprog resets the counter. This allows deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only the maximum number of frames we can place in it.
Also change the
progs/test_global_func3.cselftest to use static functions, since with the new patch it would otherwise unexpectedly pass verification.Signed-off-by Emil Tsalapatis <[email protected]>--- include/linux/bpf_verifier.h | 6 +++ kernel/bpf/verifier.c | 43 ++++++++++++------- .../selftests/bpf/progs/test_global_func3.c | 18 ++++---- 3 files changed, 42 insertions(+), 25 deletions(-)include/linux/bpf_verifier.h +2
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {/* 'start' has to be the first field otherwise find_subprog() won't work */ u32 start;/* insn idx of function entry point */
include/linux/bpf_verifier.h struct bpf_subprog_info { · line 685
Quoted 7 lines
include/linux/bpf_verifier.h +4
-0@@ -677,6 +679,10 @@ struct bpf_subprog_info {enum priv_stack_mode priv_stack_mode;struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS ]; + +int ret_insn; +int frame; +int cidx;
nit: can you add comment that cidx means caller index, i saw it below and it clarifies this code a lot.
Quoted 23 lines
include/linux/bpf_verifier.h +0
-0};struct bpf_verifier_env ;kernel/bpf/verifier.c +5
-3@@ -6652,9 +6652,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,struct bpf_insn *insn = env->prog->insnsi;int depth =0 , frame =0 , i, subprog_end, subprog_depth;bool tail_call_reachable = false; -int ret_insn[MAX_CALL_FRAMES ]; -int ret_prog[MAX_CALL_FRAMES ]; -int j; +int total; +int tmp; + +/* no caller idx */ + subprog[idx].cidx =-1 ; i = subprog[idx].start;if (!priv_stack_supported)
kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6711
Quoted 4 lines
kernel/bpf/verifier.c +1
-0@@ -6706,8 +6708,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, }else { depth += subprog_depth;if (depth >MAX_BPF_STACK ) { +for (total =1 ; subprog[idx].cidx >=0 ; total++)
nit: I think the way this loop is implemented below is more readable:
for (tmp = idx; tmp >= 0; tmp = subprog[tmp].cidx)
total++;
it is confusing now, because for loop initializes and increments total, but tracks cidx for exit check.
Quoted 158 lines
kernel/bpf/verifier.c +21
-13+ idx = subprog[idx].cidx; +verbose (env,"combined stack size of %d calls is %d. Too large\n" , - frame +1 , depth); + total, depth);return -EACCES ; } }@@ -6723,8 +6728,8 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,continue ;if (subprog[idx].is_cb) err = true; -for (int c =0 ; c < frame && !err; c++) { -if (subprog[ret_prog[c]].is_cb) { +for (tmp = idx; tmp >=0 && !err; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_cb) { err = true;break ; }@@ -6740,8 +6745,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (!bpf_pseudo_call (insn + i) && !bpf_pseudo_func (insn + i))continue ;/* remember insn and function to return to */ - ret_insn[frame] = i +1 ; - ret_prog[frame] = idx; + + subprog[idx].frame = frame; + subprog[idx].ret_insn = i +1 ;/* find the callee */ next_insn = i + insn[i].imm +1 ;@@ -6762,6 +6768,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, } } i = next_insn; + +/* caller idx */ + subprog[sidx].cidx = idx; idx = sidx;if (!priv_stack_supported) subprog[idx].priv_stack_mode =NO_PRIV_STACK ;@@ -6769,7 +6778,7 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (subprog[idx].has_tail_call) tail_call_reachable = true; - frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);@@ -6783,12 +6792,12 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,* tail call counter throughout bpf2bpf calls combined with tailcalls */ if (tail_call_reachable) -for (j =0 ; j < frame; j++) { -if (subprog[ret_prog[j]].is_exception_cb) { +for (tmp = idx; tmp >=0 ; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_exception_cb) {verbose (env,"cannot tail call within exception cb\n" );return -EINVAL ; } - subprog[ret_prog[j]].tail_call_reachable = true; + subprog[tmp].tail_call_reachable = true; }if (subprog[0 ].tail_call_reachable) env->prog->aux->tail_call_reachable = true;@@ -6796,13 +6805,15 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,/* end of for() loop means the last insn of the 'subprog' * was reached. Doesn't matter whether it was JA or EXIT */ -if (frame ==0 ) +if (frame ==0 && subprog[idx].cidx <0 )return 0 ;if (subprog[idx].priv_stack_mode !=PRIV_STACK_ADAPTIVE ) depth -=round_up_stack_depth (env, subprog[idx].stack_depth); - frame--; - i = ret_insn[frame]; - idx = ret_prog[frame]; + + idx = subprog[idx].cidx; + frame = subprog[idx].frame; + i = subprog[idx].ret_insn; + goto continue_func; }tools/testing/selftests/bpf/progs/test_global_func3.c +9
-9@@ -5,56 +5,56 @@#include <bpf/bpf_helpers.h> #include "bpf_misc.h" -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f1 (struct __sk_buff *skb) {return skb->len; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f2 (int val,struct __sk_buff *skb) {return f1 (skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f3 (int val,struct __sk_buff *skb,int var) {return f2 (var, skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f4 (struct __sk_buff *skb) {return f3 (1 , skb,2 ); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f5 (struct __sk_buff *skb) {return f4 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f6 (struct __sk_buff *skb) {return f5 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f7 (struct __sk_buff *skb) {return f6 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f8 (struct __sk_buff *skb) {return f7 (skb); } SEC("tc") -__failure __msg("the call stack of 8 frames") +__failure __msg("the call stack of 9 frames") int global_func3(struct __sk_buff *skb) {return f8 (skb);Signature · 1 line
2.49.0
Emil Tsalapatis
reviewerRe: [PATCH bpf-next v3 1/2] bpf: Only enforce 8 frame call stack limit for all-static stacks
Quoted 44 lines · Mykyta Yatsenko
Quoted 42 lines · Emil Tsalapatis
The BPF verifier currently enforces a call stack depth of 8 frames, regardless of the actual stack space consumption of those frames. The limit is necessary for static call stacks, because the bookkeeping data structures used by the verifier when stepping into static functions during verification only support 8 stack frames. However, this limitation only matters for static stack frames: Global subprogs are verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification now only fails when there is a sequence of 8 calls to non-global subprogs. Calling into a global subprog resets the counter. This allows deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only the maximum number of frames we can place in it.
Also change the
progs/test_global_func3.cselftest to use static functions, since with the new patch it would otherwise unexpectedly pass verification.Signed-off-by Emil Tsalapatis <[email protected]>--- include/linux/bpf_verifier.h | 6 +++ kernel/bpf/verifier.c | 43 ++++++++++++------- .../selftests/bpf/progs/test_global_func3.c | 18 ++++---- 3 files changed, 42 insertions(+), 25 deletions(-)include/linux/bpf_verifier.h +2
-0@@ -650,6 +650,8 @@ enum priv_stack_mode { PRIV_STACK_ADAPTIVE, }; +struct bpf_subprog_info ; +struct bpf_subprog_info {/* 'start' has to be the first field otherwise find_subprog() won't work */ u32 start;/* insn idx of function entry point */ @@ -677,6 +679,10 @@ struct bpf_subprog_info {
Quoted 9 lines
include/linux/bpf_verifier.h struct bpf_subprog_info { · line 685
Quoted 6 lines
include/linux/bpf_verifier.h +4
-0enum priv_stack_mode priv_stack_mode;struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS ]; + +int ret_insn; +int frame; +int cidx;nit: can you add comment that cidx means caller index, i saw it below and it clarifies this code a lot.
Quoted 1 line
include/linux/bpf_verifier.h +0
-0};
Ack, I will add a comment on this.
Quoted 28 lines
Quoted 22 lines
include/linux/bpf_verifier.h +0
-0struct bpf_verifier_env ;kernel/bpf/verifier.c +5
-3@@ -6652,9 +6652,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,struct bpf_insn *insn = env->prog->insnsi;int depth =0 , frame =0 , i, subprog_end, subprog_depth;bool tail_call_reachable = false; -int ret_insn[MAX_CALL_FRAMES ]; -int ret_prog[MAX_CALL_FRAMES ]; -int j; +int total; +int tmp; + +/* no caller idx */ + subprog[idx].cidx =-1 ; i = subprog[idx].start;if (!priv_stack_supported)kernel/bpf/verifier.c static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, · line 6711
Quoted 4 lines
kernel/bpf/verifier.c +1
-0@@ -6706,8 +6708,11 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, }else { depth += subprog_depth;if (depth >MAX_BPF_STACK ) { +for (total =1 ; subprog[idx].cidx >=0 ; total++)nit: I think the way this loop is implemented below is more readable:
for (tmp = idx; tmp >= 0; tmp = subprog[tmp].cidx)
Quoted 3 lines
total++;it is confusing now, because for loop initializes and increments total, but tracks cidx for exit check.
Ack, I'll update it in the respin to be more like the others.
Quoted 158 lines
Quoted 158 lines
kernel/bpf/verifier.c +21
-13+ idx = subprog[idx].cidx; +verbose (env,"combined stack size of %d calls is %d. Too large\n" , - frame +1 , depth); + total, depth);return -EACCES ; } }@@ -6723,8 +6728,8 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,continue ;if (subprog[idx].is_cb) err = true; -for (int c =0 ; c < frame && !err; c++) { -if (subprog[ret_prog[c]].is_cb) { +for (tmp = idx; tmp >=0 && !err; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_cb) { err = true;break ; }@@ -6740,8 +6745,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (!bpf_pseudo_call (insn + i) && !bpf_pseudo_func (insn + i))continue ;/* remember insn and function to return to */ - ret_insn[frame] = i +1 ; - ret_prog[frame] = idx; + + subprog[idx].frame = frame; + subprog[idx].ret_insn = i +1 ;/* find the callee */ next_insn = i + insn[i].imm +1 ;@@ -6762,6 +6768,9 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx, } } i = next_insn; + +/* caller idx */ + subprog[sidx].cidx = idx; idx = sidx;if (!priv_stack_supported) subprog[idx].priv_stack_mode =NO_PRIV_STACK ;@@ -6769,7 +6778,7 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,if (subprog[idx].has_tail_call) tail_call_reachable = true; - frame++; + frame =subprog_is_global (env, idx) ?0 : frame +1 ;if (frame >=MAX_CALL_FRAMES ) {verbose (env,"the call stack of %d frames is too deep !\n" , frame);@@ -6783,12 +6792,12 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,* tail call counter throughout bpf2bpf calls combined with tailcalls */ if (tail_call_reachable) -for (j =0 ; j < frame; j++) { -if (subprog[ret_prog[j]].is_exception_cb) { +for (tmp = idx; tmp >=0 ; tmp = subprog[tmp].cidx) { +if (subprog[tmp].is_exception_cb) {verbose (env,"cannot tail call within exception cb\n" );return -EINVAL ; } - subprog[ret_prog[j]].tail_call_reachable = true; + subprog[tmp].tail_call_reachable = true; }if (subprog[0 ].tail_call_reachable) env->prog->aux->tail_call_reachable = true;@@ -6796,13 +6805,15 @@ static int check_max_stack_depth_subprog(struct bpf_verifier_env *env, int idx,/* end of for() loop means the last insn of the 'subprog' * was reached. Doesn't matter whether it was JA or EXIT */ -if (frame ==0 ) +if (frame ==0 && subprog[idx].cidx <0 )return 0 ;if (subprog[idx].priv_stack_mode !=PRIV_STACK_ADAPTIVE ) depth -=round_up_stack_depth (env, subprog[idx].stack_depth); - frame--; - i = ret_insn[frame]; - idx = ret_prog[frame]; + + idx = subprog[idx].cidx; + frame = subprog[idx].frame; + i = subprog[idx].ret_insn; + goto continue_func; }tools/testing/selftests/bpf/progs/test_global_func3.c +9
-9@@ -5,56 +5,56 @@#include <bpf/bpf_helpers.h> #include "bpf_misc.h" -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f1 (struct __sk_buff *skb) {return skb->len; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f2 (int val,struct __sk_buff *skb) {return f1 (skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f3 (int val,struct __sk_buff *skb,int var) {return f2 (var, skb) + val; } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f4 (struct __sk_buff *skb) {return f3 (1 , skb,2 ); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f5 (struct __sk_buff *skb) {return f4 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f6 (struct __sk_buff *skb) {return f5 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f7 (struct __sk_buff *skb) {return f6 (skb); } -__attribute__ ((noinline)) +static __attribute__ ((noinline))int f8 (struct __sk_buff *skb) {return f7 (skb); } SEC("tc") -__failure __msg("the call stack of 8 frames") +__failure __msg("the call stack of 9 frames") int global_func3(struct __sk_buff *skb) {return f8 (skb);Signature · 1 line
2.49.0
Emil Tsalapatis
reviewerPATCH v3 2/2 bpf: Add deep call stack selftests
In reply to Emil TsalapatisAdd tests that demonstrate the verifier support for deep call stacks while still enforcing maximum stack size limits.
--- .../bpf/prog_tests/test_global_funcs.c | 2 + .../bpf/progs/test_global_func_deep_stack.c | 79 +++++++++++++++++++ 2 files changed, 81 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/test_global_func_deep_stack.c
tools/testing/selftests/bpf/prog_tests/test_global_funcs.c +2 -0
@@ -18,6 +18,7 @@#include "test_global_func15.skel.h" #include "test_global_func16.skel.h" #include "test_global_func17.skel.h" +#include "test_global_func_deep_stack.skel.h" #include "test_global_func_ctx_args.skel.h" #include "bpf/libbpf_internal.h"
@@ -155,6 +156,7 @@ void test_test_global_funcs(void)RUN_TESTS (test_global_func15);RUN_TESTS (test_global_func16);RUN_TESTS (test_global_func17); +RUN_TESTS (test_global_func_deep_stack);RUN_TESTS (test_global_func_ctx_args);if (test__start_subtest ("ctx_arg_rewrite" ))
tools/testing/selftests/bpf/progs/test_global_func_deep_stack.c new file +79 -0
79 lines not on this page · raw message
Signature · 1 line
2.49.0
Re: [PATCH bpf-next v3 2/2] bpf: Add deep call stack selftests
Quoted 5 lines · Emil Tsalapatis
Add tests that demonstrate the verifier support for deep call stacks while still enforcing maximum stack size limits.
Signed-off-by Emil Tsalapatis <[email protected]>---
[...]
Quoted 1 line
+/* Check we actually honor stack limits (33 * 16 = 528 > 512 =
MAX_STACK_DEPTH). */
Nit: I was wondering why stack usage per func is 16, as each one only
consumes one slot. Turns out round_up_stack_depth() rounds things up to 16. Could you please either add a comment here, or just add a second long in each function?
Quoted 6 lines
tools/testing/selftests/bpf/progs/test_global_func_deep_stack.c new file +6
-0+SEC("syscall") +__failure __msg("combined stack size of 34 calls") +int global_func_deep_stack_fail (struct __sk_buff *skb) +{ +return f32 (123 ); +}
Emil Tsalapatis
reviewerRe: [PATCH bpf-next v3 2/2] bpf: Add deep call stack selftests
Quoted 9 lines · Eduard Zingerman
Quoted 5 lines · Emil Tsalapatis
Add tests that demonstrate the verifier support for deep call stacks while still enforcing maximum stack size limits.
Signed-off-by Emil Tsalapatis <[email protected]>---
Acked-by Eduard Zingerman <[email protected]>
Quoted 9 lines
[...]
Quoted 1 line
+/* Check we actually honor stack limits (33 * 16 = 528 > 512 =
MAX_STACK_DEPTH). */Nit: I was wondering why stack usage per func is 16, as each one only
consumes one slot. Turns outround_up_stack_depth()rounds things up to 16. Could you please either add a comment here, or just add a second long in each function?
Ack, I will clarify why this is the case in a comment.
Quoted 6 lines
Quoted 6 lines
tools/testing/selftests/bpf/progs/test_global_func_deep_stack.c new file +6
-0+SEC("syscall") +__failure __msg("combined stack size of 34 calls") +int global_func_deep_stack_fail (struct __sk_buff *skb) +{ +return f32 (123 ); +}