[PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Jay Wang <hidden> · 2026-10-01
[PATCH bpf-next v4 01/12] bpf: pass the vmlinux BTF to btf_parse_module() and let it adopt the data · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 01/12] bpf: pass the vmlinux BTF to btf_parse_module() and let it adopt the data · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 02/12] bpf: split the kfunc, dtor kfunc and struct_ops registration bodies · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 02/12] bpf: split the kfunc, dtor kfunc and struct_ops registration bodies · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 03/12] bpf: fetch the vmlinux BTF where kernel types enter a program · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 03/12] bpf: fetch the vmlinux BTF where kernel types enter a program · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 04/12] bpf: take the vmlinux BTF from the btf_vmlinux module · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 04/12] bpf: take the vmlinux BTF from the btf_vmlinux module · bot+bpf-ci@kernel.org · 2026-10-01
Re: [PATCH bpf-next v4 04/12] bpf: take the vmlinux BTF from the btf_vmlinux module · sashiko-bot@kernel.org · 2026-10-02
Re: [PATCH bpf-next v4 04/12] bpf: take the vmlinux BTF from the btf_vmlinux module · Alexei Starovoitov <hidden> · 2026-10-02
[PATCH bpf-next v4 05/12] bpf, tracing: load the vmlinux BTF where tracefs and bpffs requests start · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 05/12] bpf, tracing: load the vmlinux BTF where tracefs and bpffs requests start · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 06/12] bpf: defer vmlinux kfunc and struct_ops registrations · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 06/12] bpf: defer vmlinux kfunc and struct_ops registrations · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 07/12] bpf: keep module BTF until the vmlinux BTF is available · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 07/12] bpf: keep module BTF until the vmlinux BTF is available · bot+bpf-ci@kernel.org · 2026-10-01
Re: [PATCH bpf-next v4 07/12] bpf: keep module BTF until the vmlinux BTF is available · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 08/12] bpf: expose deferred .BTF.base module BTF in sysfs from module load · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 08/12] bpf: expose deferred .BTF.base module BTF in sysfs from module load · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 09/12] bpf, trace, net: prepare CONFIG_DEBUG_INFO_BTF checks for a tristate · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 09/12] bpf, trace, net: prepare CONFIG_DEBUG_INFO_BTF checks for a tristate · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 10/12] resolve_btfids: add --btf_link to fill in .BTF.link records · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 10/12] resolve_btfids: add --btf_link to fill in .BTF.link records · bot+bpf-ci@kernel.org · 2026-10-01
Re: [PATCH bpf-next v4 10/12] resolve_btfids: add --btf_link to fill in .BTF.link records · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 11/12] tools, samples: take the vmlinux BTF from vmlinux.unstripped first · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 11/12] tools, samples: take the vmlinux BTF from vmlinux.unstripped first · sashiko-bot@kernel.org · 2026-10-02
[PATCH bpf-next v4 12/12] kbuild, bpf: allow building the vmlinux BTF as a module · Jay Wang <hidden> · 2026-10-01
Re: [PATCH bpf-next v4 12/12] kbuild, bpf: allow building the vmlinux BTF as a module · sashiko-bot@kernel.org · 2026-10-02
Re: [PATCH bpf-next v4 12/12] kbuild, bpf: allow building the vmlinux BTF as a module · Alan Maguire <hidden> · 2026-10-02
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Ihor Solodrai <ihor.solodrai@linux.dev> · 2026-10-02
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Jay Wang <hidden> · 2026-10-02
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Alan Maguire <hidden> · 2026-10-02
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Ihor Solodrai <ihor.solodrai@linux.dev> · 2026-10-02
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Alexei Starovoitov <hidden> · 2026-10-03
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Alan Maguire <hidden> · 2026-10-03
Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory · Alexei Starovoitov <hidden> · 2026-10-03

Re: [PATCH bpf-next v4 00/12] bpf: make the vmlinux BTF an on-demand loadable module (CONFIG_DEBUG_INFO_BTF=m) to save ~5.4 MB memory

From: Alan Maguire <hidden>
Date: 2026-10-03 11:46:33
Also in: bpf, linux-doc, linux-input, linux-kbuild, linux-kselftest, linux-modules, linux-perf-users, lkml, rust-for-linux, sched-ext

On Fri, Oct 02, 2026 at 01:58 PM Ihor Solodrai [off-list ref] wrote:

quoted

     Keep zstd-compressed blob in the kernel image and decompress and
     parse it synchronously on first use.

imo this is the only one of the three worth considering.

This approach is clever; my prior thinking on this was since vmlinux is often
compressed anyway we don't win anything from having a compressed .BTF section.
But as you say if we compress the .BTF section within the kernel vmlinux,
when the kernel image is then uncompressed into memory we will save most of the
space taken up by vmlinux BTF due to its still-compressed size.

The only overall cost is if when we wind up using vmlinux BTF with the compressed
.BTF section, since we have to do a full vmlinux BTF allocation (on top of the
spac taken up by the compressed .BTF section) right? So slight penalty if vmlinux
is used, but big gain if not.

If I'm understanding it correctly, seems like this is a nice win!

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help