From: Kris Van Hees <kris.van.hees@oracle.com>
To: Nick Alcock <nick.alcock@oracle.com>
Cc: dtrace@lists.linux.dev, dtrace-devel@oss.oracle.com
Subject: Re: [DTrace-devel] [PATCH v4 3/5] cg: add argument mapping in the trampoline
Date: Fri, 1 Nov 2024 20:27:18 -0400 [thread overview]
Message-ID: <ZyVx5oA+nJSqzctw@oracle.com> (raw)
In-Reply-To: <20241101155712.397674-4-nick.alcock@oracle.com>
On Fri, Nov 01, 2024 at 03:57:10PM +0000, Nick Alcock via DTrace-devel wrote:
> Before now, argument mapping was restricted to the args[] array, and was
> implemented in dt_cg_array_op in the same place where we decide whether
> to automatically copyin() userspace strings.
>
> But existing DTrace testcases suggest that arg mapping is also applied to
> argN for USDT, even though this is inconsistent with SDT and arguably less
> flexible than having argN be the unmapped arguments in all cases. Add a new
> function dt_cg_tramp_map_args(), which can be called from trampolines to
> apply mappings (destructively, but it keeps a copy of the old args in the
> save area so you can unapply them if you need to). No existing providers do
> any mapping, so this is not yet called, but it's about to be.
>
> Signed-off-by: Nick Alcock <nick.alcock@oracle.com>
> ---
> libdtrace/dt_cg.c | 36 +++++++++++++++++++++++++++++++++++-
> libdtrace/dt_cg.h | 1 +
> 2 files changed, 36 insertions(+), 1 deletion(-)
>
> diff --git a/libdtrace/dt_cg.c b/libdtrace/dt_cg.c
> index 39c27ab0ec0b..b90757bcb5cb 100644
> --- a/libdtrace/dt_cg.c
> +++ b/libdtrace/dt_cg.c
> @@ -831,6 +831,35 @@ dt_cg_tramp_restore_args(dt_pcb_t *pcb)
> }
> }
>
> +/*
> + * Move arguments according to the mappings in the array of arguments given.
> + * The original arguments are overwritten if necessary (but saved). In effect,
> + * this changes the native arguments: the caller should adjust the mappings
> + * array accordingly, and shuffle the native arg types to match.
> + *
> + * The caller must ensure that %r7 contains the value set by the
> + * dt_cg_tramp_prologue*() functions.
> + */
> +void
> +dt_cg_tramp_map_args(dt_pcb_t *pcb, dt_argdesc_t *args, size_t nargs)
> +{
> + dt_irlist_t *dlp = &pcb->pcb_ir;
> + int i;
> +
> + /*
> + * Work from saved args so we don't need to worry about overwriting regs
> + * we're about to map into different positions.
> + */
> + dt_cg_tramp_save_args(pcb);
I still maintain that this should be done in the caller. (See patch 4/5 for
more info on that.) As mentioned before, you can mention in the comment before
this function that the caller must save the arguments, just like you already
mention that %r7 must contain a value set by dt_cg_tramp_prologue*().
> +
> + for (i = 0; i < nargs; i++) {
> + if (args[i].mapping != i) {
> + emit(dlp, BPF_LOAD(BPF_DW, BPF_REG_0, BPF_REG_7, DMST_ORIG_ARG(args[i].mapping)));
> + emit(dlp, BPF_STORE(BPF_DW, BPF_REG_7, DMST_ARG(i), BPF_REG_0));
> + }
> + }
> +}
> +
> typedef struct {
> dt_irlist_t *dlp;
> dt_activity_t act;
> @@ -5060,7 +5089,12 @@ dt_cg_array_op(dt_node_t *dnp, dt_irlist_t *dlp, dt_regset_t *drp)
> * array index according to the static argument mapping (if any),
> * unless the argument reference is provided by a dynamic translator.
> * If we're using a dynamic translator for args[], then just set dn_reg
> - * to an invalid reg and return: DIF_OP_XLARG will fetch the arg later.
You dropped the end of a sentence without replacing it with anything?
> + *
> + * If dt_cg_tramp_map_args is in use, you should apply the mapping at
> + * the probe_info stage, since the effective "native" arg positions will
> + * be changed.
This does not belong here. If anything, clauses get compiled separete from
trampolines and a given clause (compiled once) may get used in more than one
BPF program, i.e. it may get associated with more than one trampoline. So
you can't really depend on what a specific trampoline does when generating code
for a clause.
> + *
> + * If this is a userland variable, note that we need to copy it in.
> */
> if (idp->di_id == DIF_VAR_ARGS) {
> if ((idp->di_kind == DT_IDENT_XLPTR ||
> diff --git a/libdtrace/dt_cg.h b/libdtrace/dt_cg.h
> index 0a7c7ba6a8c5..fb26c125adbc 100644
> --- a/libdtrace/dt_cg.h
> +++ b/libdtrace/dt_cg.h
> @@ -34,6 +34,7 @@ extern void dt_cg_tramp_get_var(dt_pcb_t *pcb, const char *name, int isstore,
> extern void dt_cg_tramp_del_var(dt_pcb_t *pcb, const char *name);
> extern void dt_cg_tramp_save_args(dt_pcb_t *pcb);
> extern void dt_cg_tramp_restore_args(dt_pcb_t *pcb);
> +extern void dt_cg_tramp_map_args(dt_pcb_t *pcb, dt_argdesc_t *args, size_t nargs);
> extern void dt_cg_tramp_call_clauses(dt_pcb_t *pcb, const dt_probe_t *prp,
> dt_activity_t act);
> extern void dt_cg_tramp_return(dt_pcb_t *pcb);
> --
> 2.46.0.278.g36e3a12567
>
>
> _______________________________________________
> DTrace-devel mailing list
> DTrace-devel@oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/dtrace-devel
next prev parent reply other threads:[~2024-11-02 0:37 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-01 15:57 [PATCH v4 0/5] usdt typed args, translators and arg remapping Nick Alcock
2024-11-01 15:57 ` [PATCH v4 1/5] usdt: get arg types and xlations into DTrace from the DOF Nick Alcock
2024-11-01 23:46 ` [DTrace-devel] " Kris Van Hees
2024-11-02 0:14 ` Kris Van Hees
2024-11-04 15:26 ` Nick Alcock
2024-11-01 15:57 ` [PATCH v4 2/5] dtprobed: stop skipping zero-tracepoint probes in dof_stash.c Nick Alcock
2024-11-01 15:57 ` [PATCH v4 3/5] cg: add argument mapping in the trampoline Nick Alcock
2024-11-02 0:27 ` Kris Van Hees [this message]
2024-11-04 16:41 ` [DTrace-devel] " Nick Alcock
2024-11-01 15:57 ` [PATCH v4 4/5] usdt: typed args and arg mapping Nick Alcock
2024-11-02 1:09 ` Kris Van Hees
2024-11-04 16:38 ` Nick Alcock
2024-11-01 15:57 ` [PATCH v4 5/5] usdt: fix create_underlying error path Nick Alcock
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ZyVx5oA+nJSqzctw@oracle.com \
--to=kris.van.hees@oracle.com \
--cc=dtrace-devel@oss.oracle.com \
--cc=dtrace@lists.linux.dev \
--cc=nick.alcock@oracle.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox