06/11/2026 10:58, David Marchand: > On Tue, 3 Nov 2026 at 18:07, Thomas Monjalon <[email protected]> wrote: > > > > Derive the hotplug devargs buffer size from the IPC parameter capacity. > > This increases the maximum devargs length in multi-process mode > > from 127 to 247 bytes, excluding the terminating null byte. > > > > Signed-off-by: Thomas Monjalon <[email protected]> > > --- > > lib/eal/common/eal_common_dev.c | 2 +- > > lib/eal/common/hotplug_mp.h | 8 +++++++- > > 2 files changed, 8 insertions(+), 2 deletions(-) > > > > diff --git a/lib/eal/common/eal_common_dev.c > > b/lib/eal/common/eal_common_dev.c > > index f42a8d2349..febb584313 100644 > > --- a/lib/eal/common/eal_common_dev.c > > +++ b/lib/eal/common/eal_common_dev.c > > @@ -276,7 +276,7 @@ rte_dev_probe(const char *devargs) > > goto skip_mp_req; > > > > if (strlen(devargs) >= EAL_DEV_MP_DEV_ARGS_MAX_LEN) { > > - EAL_LOG(ERR, "devargs truncated (len %zu, max %d)", > > + EAL_LOG(ERR, "devargs truncated (len %zu, max %zu)", > > strlen(devargs), EAL_DEV_MP_DEV_ARGS_MAX_LEN); > > return -E2BIG; > > } > > diff --git a/lib/eal/common/hotplug_mp.h b/lib/eal/common/hotplug_mp.h > > index 7221284286..bd2f0f296a 100644 > > --- a/lib/eal/common/hotplug_mp.h > > +++ b/lib/eal/common/hotplug_mp.h > > @@ -5,6 +5,7 @@ > > #ifndef _HOTPLUG_MP_H_ > > #define _HOTPLUG_MP_H_ > > > > +#include "rte_eal.h" > > #include "rte_dev.h" > > > > #define EAL_DEV_MP_ACTION_REQUEST "eal_dev_mp_request" > > @@ -12,7 +13,9 @@ > > > > #define EAL_DEV_MP_DEV_NAME_MAX_LEN RTE_DEV_NAME_MAX_LEN > > #define EAL_DEV_MP_BUS_NAME_MAX_LEN 32 > > -#define EAL_DEV_MP_DEV_ARGS_MAX_LEN 128 > > +#define EAL_DEV_MP_DEV_ARGS_MAX_LEN (RTE_MP_MAX_PARAM_LEN \ > > + - sizeof(enum eal_dev_req_type) - sizeof(int)) > > + /* minus t and result from struct eal_dev_mp_req */ > > > > enum eal_dev_req_type { > > EAL_DEV_REQ_TYPE_ATTACH, > > @@ -27,6 +30,9 @@ struct eal_dev_mp_req { > > int result; > > }; > > > > +static_assert(sizeof(struct eal_dev_mp_req) <= RTE_MP_MAX_PARAM_LEN, > > + "Hotplug message struct exceeds IPC parameter capacity"); > > + > > A bit ugly to list those fields, even with the assert. > > No strong opinion, just listing an alternative. > > The idea is to switch to a flexible array for devargs. > > struct eal_dev_mp_req { > enum eal_dev_req_type t; > - char devargs[EAL_DEV_MP_DEV_ARGS_MAX_LEN]; > int result; > + char devargs[]; > }; > > +#define EAL_DEV_MP_DEV_ARGS_MAX_LEN \ > + (RTE_MP_MAX_PARAM_LEN - sizeof(struct eal_dev_mp_req)) > > > This requires an update of > eal_dev_hotplug_request_to_primary/eal_dev_hotplug_request_to_secondary > so they manipulate a rte_mp_msg object. > A bonus side effect is that it avoids copies on stack for the message, > as the caller already provides the message buffer.
Yes it looks better.
