From: Johannes Zink <j.zink@pengutronix.de>
To: Xu Yilun <yilun.xu@intel.com>
Cc: devicetree@vger.kernel.org, linux-fpga@vger.kernel.org,
Rob Herring <robh+dt@kernel.org>, Moritz Fischer <mdf@kernel.org>,
kernel@pengutronix.de, Wu Hao <hao.wu@intel.com>
Subject: Re: [PATCH 07/16] fpga: machxo2-spi: fix big-endianness incompatibility
Date: Mon, 29 Aug 2022 12:41:03 +0200 [thread overview]
Message-ID: <b8a6c62facf182b9ad4f333af69072e1a8c0293b.camel@pengutronix.de> (raw)
In-Reply-To: <Ywx2eeTvBhNw62pW@yilunxu-OptiPlex-7050>
Hi Yilun,
On Mon, 2022-08-29 at 16:19 +0800, Xu Yilun wrote:
> On 2022-08-25 at 16:13:34 +0200, Johannes Zink wrote:
> > The SPI message is written into the lowest-addressed bits of an
> > unsigned long variable, but be32_to_cpu is called on the least
> > significant bits of the variable. On big-endian 64-bit systems,
> > this would give a wrong result. Fix this by using the fixed-size
> > u32 instead of unsigned long. This clashes with the use of
> > test_bit, which is unnecessary for a single u32 variable, so
> > we adjust all usage sites appropriately and prefix the macros
> > with MACHXO2_ while at it.
> >
> > Signed-off-by: Johannes Zink <j.zink@pengutronix.de>
> > ---
> > drivers/fpga/machxo2-spi.c | 110 ++++++++++++++++++---------------
> > ----
> > 1 file changed, 55 insertions(+), 55 deletions(-)
> >
> > diff --git a/drivers/fpga/machxo2-spi.c b/drivers/fpga/machxo2-
> > spi.c
> > index 5e12612c7289..d1a8f28e04e7 100644
> > --- a/drivers/fpga/machxo2-spi.c
> > +++ b/drivers/fpga/machxo2-spi.c
> > @@ -9,6 +9,7 @@
> > */
> >
> > #include <linux/delay.h>
> > +#include <linux/bitfield.h>
> > #include <linux/fpga/fpga-mgr.h>
> > #include <linux/gpio/consumer.h>
> > #include <linux/module.h>
> > @@ -41,41 +42,40 @@
> > #define MACHXO2_BUF_SIZE (MACHXO2_PAGE_SIZE + 4)
> >
> > /* Status register bits, errors and error mask */
> > -#define BUSY 12
> > -#define DONE 8
> > -#define DVER 27
> > -#define ENAB 9
> > -#define ERRBITS 23
> > -#define ERRMASK 7
> > -#define FAIL 13
> > -
> > -#define ENOERR 0 /* no error */
> > -#define EID 1
> > -#define ECMD 2
> > -#define ECRC 3
> > -#define EPREAM 4 /* preamble error */
> > -#define EABRT 5 /* abort error */
> > -#define EOVERFL 6 /* overflow error */
> > -#define ESDMEOF 7 /* SDM EOF */
> > -
> > -static inline u8 get_err(unsigned long *status)
> > +#define MACHXO2_BUSY BIT(12)
> > +#define MACHXO2_DONE BIT(8)
> > +#define MACHXO2_DVER BIT(27)
> > +#define MACHXO2_ENAB BIT(9)
> > +#define MACHXO2_ERR GENMASK(25, 23)
> > +#define MACHXO2_ERR_ENOERR 0 /* no error */
> > +#define MACHXO2_ERR_EID 1
> > +#define MACHXO2_ERR_ECMD 2
> > +#define MACHXO2_ERR_ECRC 3
> > +#define MACHXO2_ERR_EPREAM 4 /* preamble error */
> > +#define MACHXO2_ERR_EABRT 5 /* abort error */
> > +#define MACHXO2_ERR_EOVERFL 6 /* overflow error */
> > +#define MACHXO2_ERR_ESDMEOF 7 /* SDM EOF */
> > +#define MACHXO2_FAIL BIT(13)
> > +
> > +static inline u8 get_err(u32 status)
> > {
> > - return (*status >> ERRBITS) & ERRMASK;
> > + return FIELD_GET(MACHXO2_ERR, status);
> > }
>
> So far the changes have nothing to do with the endian issue, is it?
> So
> please put it into another patch.
This is explained in the commit message. We need to get rid of unsigned
long, but test_bit requires using that type. Using the BIT and GENMASK
macros instead allows us to work on a u32 directly.
Admittedly, prefixing the macros with MACHXO2_ can be separated out,
but as I am already touching them to make them into masks, which is
required when not using test_bit, I chose to squash this here.
This is also mentioned in the commit message. If you prefer, I do
the rename in a separate patch. I can do that for v2, but I don't
see how I can cleanly separate s/unsigned long/u32/ and test_bit
removal.
>
> >
> > -static int get_status(struct spi_device *spi, unsigned long
> > *status)
> > +static int get_status(struct spi_device *spi, u32 *status)
> > {
> > struct spi_message msg;
> > struct spi_transfer rx, tx;
> > static const u8 cmd[] = LSC_READ_STATUS;
> > + __be32 tmp;
> > int ret;
> >
> > memset(&rx, 0, sizeof(rx));
> > memset(&tx, 0, sizeof(tx));
> > tx.tx_buf = cmd;
> > tx.len = sizeof(cmd);
> > - rx.rx_buf = status;
> > - rx.len = 4;
> > + rx.rx_buf = &tmp;
> > + rx.len = sizeof(tmp);
> > spi_message_init(&msg);
> > spi_message_add_tail(&tx, &msg);
> > spi_message_add_tail(&rx, &msg);
> > @@ -83,7 +83,7 @@ static int get_status(struct spi_device *spi,
> > unsigned long *status)
> > if (ret)
> > return ret;
> >
> > - *status = be32_to_cpu(*status);
> > + *status = be32_to_cpu(tmp);
>
> Why not directly operate on the status, I don't see the difference.
I wanted to state more clearly that the value read in the rx message is
big endian and needs conversion.
Besides, since be32_to_cpu takes a __be32 as an argument, passing
status, which is an u32, should give a static analyzer warning.
>
> >
> > return 0;
> > }
> > @@ -91,30 +91,30 @@ static int get_status(struct spi_device *spi,
> > unsigned long *status)
> > static const char *get_err_string(u8 err)
> > {
> > switch (err) {
> > - case ENOERR: return "No Error";
> > - case EID: return "ID ERR";
> > - case ECMD: return "CMD ERR";
> > - case ECRC: return "CRC ERR";
> > - case EPREAM: return "Preamble ERR";
> > - case EABRT: return "Abort ERR";
> > - case EOVERFL: return "Overflow ERR";
> > - case ESDMEOF: return "SDM EOF";
> > + case MACHXO2_ERR_ENOERR: return "No Error";
> > + case MACHXO2_ERR_EID: return "ID ERR";
> > + case MACHXO2_ERR_ECMD: return "CMD ERR";
> > + case MACHXO2_ERR_ECRC: return "CRC ERR";
> > + case MACHXO2_ERR_EPREAM: return "Preamble ERR";
> > + case MACHXO2_ERR_EABRT: return "Abort ERR";
> > + case MACHXO2_ERR_EOVERFL: return "Overflow ERR";
> > + case MACHXO2_ERR_ESDMEOF: return "SDM EOF";
> > }
>
> Same concern, no relation to the commit message
please see my comment above. If you prefer, I split it out in a
separate patch.
>
> >
> > - return "Default switch case";
> > + return "Unknown";
> > }
> >
> > -static void dump_status_reg(unsigned long *status)
> > +static void dump_status_reg(u32 status)
> > {
> > - pr_debug("machxo2 status: 0x%08lX - done=%d, cfgena=%d,
> > busy=%d, fail=%d, devver=%d, err=%s\n",
> > - *status, test_bit(DONE, status), test_bit(ENAB,
> > status),
> > - test_bit(BUSY, status), test_bit(FAIL, status),
> > - test_bit(DVER, status),
> > get_err_string(get_err(status)));
> > + pr_debug("machxo2 status: 0x%08X - done=%d, cfgena=%d,
> > busy=%d, fail=%d, devver=%d, err=%s\n",
> > + status, !!FIELD_GET(MACHXO2_DONE, status),
> > !!FIELD_GET(MACHXO2_ENAB, status),
> > + !!FIELD_GET(MACHXO2_BUSY, status),
> > !!FIELD_GET(MACHXO2_FAIL, status),
> > + !!FIELD_GET(MACHXO2_DVER, status),
> > get_err_string(get_err(status)));
>
> Same concern. I'll not point out one by one below.
>
> Thanks,
> Yilun
>
as I explained in the commit message, I am removing test_bit. As far as
the macro renaming is concerned, please see my comments above.
Best regards
Johannes
> > }
>
>
> >
> > static int wait_until_not_busy(struct spi_device *spi)
> > {
> > - unsigned long status;
> > + u32 status;
> > int ret, loop = 0;
> >
> > do {
> > @@ -123,7 +123,7 @@ static int wait_until_not_busy(struct
> > spi_device *spi)
> > return ret;
> > if (++loop >= MACHXO2_MAX_BUSY_LOOP)
> > return -EBUSY;
> > - } while (test_bit(BUSY, &status));
> > + } while (status & MACHXO2_BUSY);
> >
> > return 0;
> > }
> > @@ -169,14 +169,14 @@ static int machxo2_cleanup(struct
> > fpga_manager *mgr)
> >
> > static bool machxo2_status_done(unsigned long status)
> > {
> > - return !test_bit(BUSY, &status) && test_bit(DONE, &status)
> > &&
> > - get_err(&status) == ENOERR;
> > + return (((status & (MACHXO2_BUSY | MACHXO2_DONE)) ==
> > MACHXO2_DONE) &&
> > + get_err(status) == MACHXO2_ERR_ENOERR);
> > }
> >
> > static enum fpga_mgr_states machxo2_spi_state(struct fpga_manager
> > *mgr)
> > {
> > struct spi_device *spi = mgr->priv;
> > - unsigned long status;
> > + u32 status;
> >
> > get_status(spi, &status);
> > if (machxo2_status_done(status))
> > @@ -195,7 +195,7 @@ static int machxo2_write_init(struct
> > fpga_manager *mgr,
> > static const u8 enable[] = ISC_ENABLE;
> > static const u8 erase[] = ISC_ERASE;
> > static const u8 initaddr[] = LSC_INITADDRESS;
> > - unsigned long status;
> > + u32 status;
> > int ret;
> >
> > if ((info->flags & FPGA_MGR_PARTIAL_RECONFIG)) {
> > @@ -205,7 +205,7 @@ static int machxo2_write_init(struct
> > fpga_manager *mgr,
> > }
> >
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> > memset(tx, 0, sizeof(tx));
> > spi_message_init(&msg);
> > tx[0].tx_buf = &enable;
> > @@ -226,11 +226,11 @@ static int machxo2_write_init(struct
> > fpga_manager *mgr,
> > goto fail;
> >
> > get_status(spi, &status);
> > - if (test_bit(FAIL, &status)) {
> > + if (status & MACHXO2_FAIL) {
> > ret = -EINVAL;
> > goto fail;
> > }
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> >
> > spi_message_init(&msg);
> > tx[2].tx_buf = &initaddr;
> > @@ -241,7 +241,7 @@ static int machxo2_write_init(struct
> > fpga_manager *mgr,
> > goto fail;
> >
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> >
> > return 0;
> > fail:
> > @@ -258,7 +258,7 @@ static int machxo2_write(struct fpga_manager
> > *mgr, const char *buf,
> > struct spi_transfer tx;
> > static const u8 progincr[] = LSC_PROGINCRNV;
> > u8 payload[MACHXO2_BUF_SIZE];
> > - unsigned long status;
> > + u32 status;
> > int i, ret;
> >
> > if (count % MACHXO2_PAGE_SIZE != 0) {
> > @@ -266,7 +266,7 @@ static int machxo2_write(struct fpga_manager
> > *mgr, const char *buf,
> > return -EINVAL;
> > }
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> > memcpy(payload, &progincr, sizeof(progincr));
> > for (i = 0; i < count; i += MACHXO2_PAGE_SIZE) {
> > memcpy(&payload[sizeof(progincr)], &buf[i],
> > MACHXO2_PAGE_SIZE);
> > @@ -284,7 +284,7 @@ static int machxo2_write(struct fpga_manager
> > *mgr, const char *buf,
> > }
> > }
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> >
> > return 0;
> > }
> > @@ -297,7 +297,7 @@ static int machxo2_write_complete(struct
> > fpga_manager *mgr,
> > struct spi_transfer tx[2];
> > static const u8 progdone[] = ISC_PROGRAMDONE;
> > static const u8 refresh[] = LSC_REFRESH;
> > - unsigned long status;
> > + u32 status;
> > int ret, refreshloop = 0;
> >
> > memset(tx, 0, sizeof(tx));
> > @@ -313,8 +313,8 @@ static int machxo2_write_complete(struct
> > fpga_manager *mgr,
> > goto fail;
> >
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > - if (!test_bit(DONE, &status)) {
> > + dump_status_reg(status);
> > + if (!(status & MACHXO2_DONE)) {
> > machxo2_cleanup(mgr);
> > ret = -EINVAL;
> > goto fail;
> > @@ -333,7 +333,7 @@ static int machxo2_write_complete(struct
> > fpga_manager *mgr,
> >
> > /* check refresh status */
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> > if (machxo2_status_done(status))
> > break;
> > if (++refreshloop == MACHXO2_MAX_REFRESH_LOOP) {
> > @@ -344,7 +344,7 @@ static int machxo2_write_complete(struct
> > fpga_manager *mgr,
> > } while (1);
> >
> > get_status(spi, &status);
> > - dump_status_reg(&status);
> > + dump_status_reg(status);
> >
> > return 0;
> > fail:
> > --
> > 2.30.2
> >
>
>
--
Pengutronix e.K. | Johannes Zink |
Steuerwalder Str. 21 | https://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686| Fax: +49-5121-206917-5555 |
next prev parent reply other threads:[~2022-08-29 10:43 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-25 14:13 [PATCH 00/16] Add support for Lattice MachXO2 programming via I2C Johannes Zink
2022-08-25 14:13 ` [PATCH 01/16] dt-bindings: fpga: convert Lattice MachXO2 Slave binding to YAML Johannes Zink
2022-08-30 20:30 ` Rob Herring
2022-08-31 7:12 ` Johannes Zink
2022-08-25 14:13 ` [PATCH 02/16] dt-bindings: fpga: machxo2-slave: add erasure properties Johannes Zink
2022-08-29 7:39 ` Xu Yilun
[not found] ` <9d5512768acb4d57f339942007402a9ed9483e84.camel@pengutronix.de>
[not found] ` <YwzWt8KjyfdyqehI@yilunxu-OptiPlex-7050>
2022-08-31 7:38 ` Johannes Zink
2022-09-03 14:49 ` Xu Yilun
2022-08-30 20:36 ` Rob Herring
2022-08-31 7:07 ` Johannes Zink
2022-08-25 14:13 ` [PATCH 03/16] dt-bindings: fpga: machxo2-slave: add pin for program sequence init Johannes Zink
2022-08-25 18:51 ` Rob Herring
2022-08-26 7:56 ` Johannes Zink
2022-08-29 7:45 ` Xu Yilun
[not found] ` <a42d72cd71c96ca675f5bb0cf59128c7f1cb04bb.camel@pengutronix.de>
[not found] ` <YwzZYM6GU0GiqBiq@yilunxu-OptiPlex-7050>
2022-08-31 7:51 ` Johannes Zink
2022-08-31 8:08 ` Johannes Zink
2022-08-25 14:13 ` [PATCH 04/16] dt-bindings: fpga: machxo2-slave: add lattice,machxo2-slave-i2c compatible Johannes Zink
2022-08-30 20:40 ` Rob Herring
2022-08-31 7:10 ` Johannes Zink
2022-08-25 14:13 ` [PATCH 05/16] fpga: machxo2-spi: remove #ifdef DEBUG Johannes Zink
2022-08-25 14:13 ` [PATCH 06/16] fpga: machxo2-spi: factor out status check for readability Johannes Zink
2022-08-25 14:13 ` [PATCH 07/16] fpga: machxo2-spi: fix big-endianness incompatibility Johannes Zink
2022-08-29 8:19 ` Xu Yilun
2022-08-29 10:41 ` Johannes Zink [this message]
2022-08-25 14:13 ` [PATCH 08/16] fpga: machxo2-spi: simplify with spi_sync_transfer() Johannes Zink
2022-08-25 14:13 ` [PATCH 09/16] fpga: machxo2-spi: simplify spi write commands Johannes Zink
2022-08-25 14:13 ` [PATCH 10/16] fpga: machxo2-spi: prepare extraction of common code Johannes Zink
2022-08-25 14:13 ` [PATCH 11/16] fpga: machxo2: move non-spi-related functionality to " Johannes Zink
2022-08-25 14:13 ` [PATCH 12/16] fpga: machxo2: improve status register dump Johannes Zink
2022-08-25 14:13 ` [PATCH 13/16] fpga: machxo2: add optional additional flash areas to be erased Johannes Zink
2022-08-25 14:13 ` [PATCH 14/16] fpga: machxo2: add program initialization signalling via gpio Johannes Zink
2022-08-25 14:13 ` [PATCH 15/16] fpga: machxo2: extend erase timeout for machxo2 FPGA Johannes Zink
2022-08-29 9:26 ` Xu Yilun
2022-08-29 10:51 ` Johannes Zink
2022-08-29 14:57 ` Xu Yilun
2022-08-31 7:56 ` Johannes Zink
2022-08-25 14:13 ` [PATCH 16/16] fpga: machxo2: add configuration over i2c Johannes Zink
2022-08-29 9:47 ` Xu Yilun
2022-08-29 13:21 ` Johannes Zink
2022-08-29 14:45 ` Xu Yilun
2022-08-31 16:07 ` Johannes Zink
2022-08-25 15:25 ` [PATCH 00/16] Add support for Lattice MachXO2 programming via I2C Ivan Bornyakov
2022-08-26 6:32 ` Johannes Zink
2022-08-26 8:15 ` Ivan Bornyakov
2022-08-26 8:25 ` Sascha Hauer
2022-08-26 9:00 ` Ivan Bornyakov
2022-08-26 9:19 ` Ivan Bornyakov
2022-08-26 15:26 ` Xu Yilun
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=b8a6c62facf182b9ad4f333af69072e1a8c0293b.camel@pengutronix.de \
--to=j.zink@pengutronix.de \
--cc=devicetree@vger.kernel.org \
--cc=hao.wu@intel.com \
--cc=kernel@pengutronix.de \
--cc=linux-fpga@vger.kernel.org \
--cc=mdf@kernel.org \
--cc=robh+dt@kernel.org \
--cc=yilun.xu@intel.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;
as well as URLs for NNTP newsgroup(s).