Re: [PATCH v4] vhost: fix crash caused by accessing a freed vsocket

[Stephen Hemminger <stephen@networkplumber.org>]

On Mon,  8 Jul 2024 12:41:03 +0800
Gongming Chen <chengongming1900@outlook.com> wrote:

> diff --git a/lib/vhost/vhost_thread.c b/lib/vhost/vhost_thread.c
> new file mode 100644
> index 0000000000..f3ff182976
> --- /dev/null
> +++ b/lib/vhost/vhost_thread.c
> @@ -0,0 +1,33 @@
> +#include <rte_rwlock.h>
> +
> +#include "vhost_thread.h"
> +
> +static rte_rwlock_t vhost_thread_lock = RTE_RWLOCK_INITIALIZER;
> +
> +void
> +vhost_thread_read_lock(void)
> +	__rte_no_thread_safety_analysis
> +{
> +	rte_rwlock_read_lock(&vhost_thread_lock);
> +}
> +
> +void
> +vhost_thread_read_unlock(void)
> +	__rte_no_thread_safety_analysis
> +{
> +	rte_rwlock_read_unlock(&vhost_thread_lock);
> +}
> +
> +void
> +vhost_thread_write_lock(void)
> +	__rte_no_thread_safety_analysis
> +{
> +	rte_rwlock_write_lock(&vhost_thread_lock);
> +}
> +
> +void
> +vhost_thread_write_unlock(void)
> +	__rte_no_thread_safety_analysis
> +{
> +	rte_rwlock_write_unlock(&vhost_thread_lock);
> +}

Looking at the back catalog and spotted this.

What is the reason for moving all the vhost_thread locks into a file?
Also turning off thread safety is a bad idea.
And mixing DPDK locks with pthread mutex is not wise either.

This is not going to be critical path so why not use a pthread_mutex?