From: Johan Hovold <johan@kernel.org>
To: "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
Cc: Alex Elder <elder@kernel.org>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
greybus-dev@lists.linaro.org, linux-staging@lists.linux.dev,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4] staging: greybus: Convert uart.c from IDR to XArray
Date: Mon, 30 Aug 2021 11:12:28 +0200 [thread overview]
Message-ID: <YSyg/Db1So0LDGR+@hovoldconsulting.com> (raw)
In-Reply-To: <20210829092250.25379-1-fmdefrancesco@gmail.com>
On Sun, Aug 29, 2021 at 11:22:50AM +0200, Fabio M. De Francesco wrote:
> Convert greybus/uart.c from IDR to XArray. The abstract data type XArray
> is more memory-efficient, parallelisable, and cache friendly. It takes
> advantage of RCU to perform lookups without locking. Furthermore, IDR is
> deprecated because XArray has a better (cleaner and more consistent) API.
Where does it say that IDR is deprecated? Almost all drivers use IDR/IDA
and its interfaces are straight-forward. In most cases we don't care
about a possible slight increase in efficiency either, and so also in
this case. Correctness is what matters and doing these conversions risks
introducing regressions.
And I believe IDR use XArray internally these days anyway.
> Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
> ---
>
> v3->v4:
> Remove mutex_lock/unlock around xa_load(). These locks seem to
> be unnecessary because there is a 1:1 correspondence between
> a specific minor and its gb_tty and there is no reference
> counting. I think that the RCU locks used inside xa_load()
> are sufficient to protect this API from returning an invalid
> gb_tty in case of concurrent access. Some more considerations
> on this topic are in the following message to linux-kernel list:
> https://lore.kernel.org/lkml/3554184.2JXonMZcNW@localhost.localdomain/
This just doesn't make sense (and a valid motivation would need to go in
the commit message if there was one).
> v2->v3:
> Fix some issues according to a review by Alex Elder <elder@ieee.org>
> v1->v2:
> Fix an issue found by the kernel test robot. It is due to
> passing to xa_*lock() the same old mutex that IDR used with
> the previous version of the code.
>
> drivers/staging/greybus/uart.c | 32 ++++++++++++++------------------
> 1 file changed, 14 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/staging/greybus/uart.c b/drivers/staging/greybus/uart.c
> index 73f01ed1e5b7..f66983adb51b 100644
> --- a/drivers/staging/greybus/uart.c
> +++ b/drivers/staging/greybus/uart.c
> @@ -22,7 +22,7 @@
> #include <linux/serial.h>
> #include <linux/tty_driver.h>
> #include <linux/tty_flip.h>
> -#include <linux/idr.h>
> +#include <linux/xarray.h>
> #include <linux/fs.h>
> #include <linux/kdev_t.h>
> #include <linux/kfifo.h>
> @@ -32,8 +32,9 @@
>
> #include "gbphy.h"
>
> -#define GB_NUM_MINORS 16 /* 16 is more than enough */
> -#define GB_NAME "ttyGB"
> +#define GB_NUM_MINORS 16 /* 16 is more than enough */
> +#define GB_RANGE_MINORS XA_LIMIT(0, GB_NUM_MINORS)
> +#define GB_NAME "ttyGB"
>
> #define GB_UART_WRITE_FIFO_SIZE PAGE_SIZE
> #define GB_UART_WRITE_ROOM_MARGIN 1 /* leave some space in fifo */
> @@ -67,8 +68,7 @@ struct gb_tty {
> };
>
> static struct tty_driver *gb_tty_driver;
> -static DEFINE_IDR(tty_minors);
> -static DEFINE_MUTEX(table_lock);
> +static DEFINE_XARRAY(tty_minors);
>
> static int gb_uart_receive_data_handler(struct gb_operation *op)
> {
> @@ -341,8 +341,7 @@ static struct gb_tty *get_gb_by_minor(unsigned int minor)
> {
> struct gb_tty *gb_tty;
>
> - mutex_lock(&table_lock);
> - gb_tty = idr_find(&tty_minors, minor);
> + gb_tty = xa_load(&tty_minors, minor);
> if (gb_tty) {
> mutex_lock(&gb_tty->mutex);
> if (gb_tty->disconnected) {
> @@ -353,19 +352,18 @@ static struct gb_tty *get_gb_by_minor(unsigned int minor)
> mutex_unlock(&gb_tty->mutex);
> }
> }
> - mutex_unlock(&table_lock);
You can't just drop the locking here since you'd introduce a potential
use-after-free in case gb_tty is freed after the lookup but before the
port reference is taken.
That said, this driver is already broken since it can currently free the
gb_tty while there are references to the port. I'll try to fix it up...
> return gb_tty;
> }
But as you may have gathered, I don't think doing these conversions is a
good idea.
Johan
next prev parent reply other threads:[~2021-08-30 9:12 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-29 9:22 [PATCH v4] staging: greybus: Convert uart.c from IDR to XArray Fabio M. De Francesco
2021-08-30 9:12 ` Johan Hovold [this message]
2021-08-30 11:10 ` Fabio M. De Francesco
2021-08-30 11:52 ` Johan Hovold
2021-08-30 12:16 ` Matthew Wilcox
2021-08-30 12:33 ` Johan Hovold
2021-08-30 13:16 ` Fabio M. De Francesco
2021-08-30 13:20 ` [greybus-dev] " Alex Elder
2021-08-31 8:07 ` Johan Hovold
2021-08-31 10:42 ` Alex Elder
2021-08-31 11:51 ` Johan Hovold
2021-08-31 11:50 ` Fabio M. De Francesco
2021-08-31 12:18 ` Johan Hovold
2021-09-01 12:09 ` Alex Elder
2021-09-01 13:56 ` Fabio M. De Francesco
2021-09-01 14:29 ` Matthew Wilcox
2021-09-01 15:39 ` Fabio M. De Francesco
2021-08-30 13:31 ` Matthew Wilcox
2021-08-31 8:16 ` Johan Hovold
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=YSyg/Db1So0LDGR+@hovoldconsulting.com \
--to=johan@kernel.org \
--cc=elder@kernel.org \
--cc=fmdefrancesco@gmail.com \
--cc=gregkh@linuxfoundation.org \
--cc=greybus-dev@lists.linaro.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-staging@lists.linux.dev \
/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;
as well as URLs for NNTP newsgroup(s).