RFC ptrace: introduce task_struct->ptrace_cxt
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v3 RFC Latest12 patches 1 replyIn review
- v2 RFC12 patches 15 repliesSuperseded
Changes from the cover letter · 43 lines
04/12: move kfree() outside of task_lock() in alloc_ptrace_context().
But I misread your sugesstion:
> task_lock(tsk);
> if (likely(!tsk->ptrace_ctx)) {
> tsk->ptrace_ctx = ptrace_ctx;
> task_unlock(tsk);
> return 0;
> }
>
> task_unlock(tsk);
> kfree(ptrace_ctx);
Just can't do that. I hate multiple unlocks very much. So it
becomes:
task_lock(tsk);
if (likely(!tsk->ptrace_ctx)) {
tsk->ptrace_ctx = ptrace_ctx;
ptrace_ctx = NULL;
}
task_unlock(tsk);
kfree(ptrace_ctx);
kfree(NULL) is specially allowed.
05/12: Add the comment about tracehook_init_task to ptrace_init_task().
I didn't change the original comment, but imho it is confusing.
"immediately after adding @child to its parent's children list"
does not matter. What does matter, is that it is called before
this child is visible to the user-space and thus the unconditional
ptrace_link() is safe: nobody could attach before us.
08/12: No changes, but I think it better to change format_mca_init_stack()
right now. Imho it has nothing to do with tracehooks and it is the
last user of ->parent (except perhaps there are some in arch/ code
which should be tracehookfied).
09/12: Remove the stale "pt_" from the changelog.Patches in v3 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/12 | ptrace: ptrace_signal: fix the usage of ->parentkernel/ | +13 | 0 replies | |
| 2/12 | ptrace: tracehooks: do not use task->ptrace directlyinclude/ | +9 | 0 replies | |
| 3/12 | ptrace: avoid using task->ptrace when possible in ptrace.{c,h}include/ kernel/ | +31 | 0 replies | |
| 4/12 | ptrace: introduce the empty "struct ptrace_context"include/ include/ include/ +2 more | +80 | 0 replies | |
| 5/12 | ptrace: kill the now unneeded args from tracehook_finish_clone() pathinclude/ include/ kernel/ | +7 | 0 replies | |
| 6/12 | ptrace: mv task_struct->ptrace ptrace_ctx->flags, kill ptrace_link()include/ include/ kernel/ | +17 | 0 replies | |
| 7/12 | ptrace: change __ptrace_link() to set ->flags, rename to ptrace_link()include/ kernel/ | +16 | 0 replies | |
| 8/12 | ptrace: mv task->parent ptrace_ctx->tracerarch/ include/ include/ +5 more | +32 | 0 replies | |
| 9/12 | ptrace: introduce ptrace_tracer() helperinclude/ kernel/ kernel/ | +22 | 0 replies | |
| 10/12 | ptrace: mv task_struct->last_siginfo ptrace_ctx->infopinclude/ include/ kernel/ +1 more | +9 | 0 replies | |
| 11/12 | ptrace: mv task_struct->ptrace_message ptrace_ctx->messageinclude/ include/ kernel/ | +6 | 0 replies | |
| 12/12 | ptrace: mv task_struct->ptrace_entry ptrace_ctx->entryinclude/ include/ include/ +2 more | +19 | 0 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v3 yet.
Cover letter [RFC PATCH 00/12 v3] ptrace: introduce task_struct->ptrace_cxt
Changes:
04/12: move kfree() outside of task_lock() in alloc_ptrace_context().
But I misread your sugesstion:
> task_lock(tsk);
> if (likely(!tsk->ptrace_ctx)) {
> tsk->ptrace_ctx = ptrace_ctx;
> task_unlock(tsk);
> return 0;
> }
>
> task_unlock(tsk);
> kfree(ptrace_ctx);
Just can't do that. I hate multiple unlocks very much. So it
becomes:
task_lock(tsk);
if (likely(!tsk->ptrace_ctx)) {
tsk->ptrace_ctx = ptrace_ctx;
ptrace_ctx = NULL;
}
task_unlock(tsk);
kfree(ptrace_ctx);
kfree(NULL) is specially allowed.
05/12: Add the comment about tracehook_init_task to ptrace_init_task().
I didn't change the original comment, but imho it is confusing.
"immediately after adding @child to its parent's children list"
does not matter. What does matter, is that it is called before
this child is visible to the user-space and thus the unconditional
ptrace_link() is safe: nobody could attach before us.
08/12: No changes, but I think it better to change format_mca_init_stack()
right now. Imho it has nothing to do with tracehooks and it is the
last user of ->parent (except perhaps there are some in arch/ code
which should be tracehookfied).
09/12: Remove the stale "pt_" from the changelog.Oleg.