From mboxrd@z Thu Jan 1 00:00:00 1970 From: mike.looijmans@topic.nl (Mike Looijmans) Date: Tue, 20 Oct 2015 08:57:08 +0200 Subject: [PATCH] fpga: zynq-fpga: Change fw format to handle bin instead of bit. In-Reply-To: <1445299319-14921-2-git-send-email-moritz.fischer@ettus.com> References: <1445299319-14921-1-git-send-email-moritz.fischer@ettus.com> <1445299319-14921-2-git-send-email-moritz.fischer@ettus.com> Message-ID: <5625E5C4.7090803@topic.nl> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 20-10-15 02:01, Moritz Fischer wrote: > This gets rid of the code to strip away the header and byteswap. > > Signed-off-by: Moritz Fischer > --- > drivers/fpga/zynq-fpga.c | 23 ++++++++++------------- > 1 file changed, 10 insertions(+), 13 deletions(-) > > diff --git a/drivers/fpga/zynq-fpga.c b/drivers/fpga/zynq-fpga.c > index 617d382..916c551 100644 > --- a/drivers/fpga/zynq-fpga.c > +++ b/drivers/fpga/zynq-fpga.c > @@ -104,6 +104,8 @@ > #define INIT_POLL_TIMEOUT 2500000 > /* Delay for polling reset bits */ > #define INIT_POLL_DELAY 20 > +/* Sync sequence for firmware */ > +#define SYNC_SEQUENCE "\x66\x55\x99\xAA" > > /* Masks for controlling stuff in SLCR */ > /* Disable all Level shifters */ > @@ -301,24 +303,19 @@ static int zynq_fpga_ops_write(struct fpga_manager *mgr, > > memcpy(kbuf, buf, count); > > - /* look for the sync word */ > + /* look for the sync sequence */ > for (i = 0; i < count - 4; i++) { > - if (memcmp(kbuf + i, "\xAA\x99\x55\x66", 4) == 0) { > - dev_dbg(priv->dev, "Found swapped sync word\n"); > + if (memcmp(kbuf + i, SYNC_SEQUENCE, 4) == 0) { > + dev_dbg(priv->dev, "Found sync sequence"); > break; > } > } Regarding this check, and judging from the code, I'd say that any file with a non word-aligned sync word is broken. So you could speed this up considerably by walking through the array word-by-word instread of byte-by-byte. Checking the whole 4MB buffer from beginning to end doesn't make it really efficient either, if you're going to visit all that memory in case the stream is invalid, you might as well send it to the hardware, I wouldn't be surprised if that even takes less effort than going through it byte-by-byte. I for one would rather see this check removed, the hardware handles it just fine and also catches more likely errors (e.g. sending a 7030 bitstream to a 7015 chip is something that the driver cannot catch anyway, but the hardware detects that error just fine). > - /* remove the header, align the data on word boundary */ > - if (i != count - 4) { > - count -= i; > - memmove(kbuf, kbuf + i, count); > - } > - > - /* fixup endianness of the data */ > - for (i = 0; i < count; i += 4) { > - u32 *p = (u32 *)&kbuf[i]; > - *p = swab32(*p); > + /* if we detect something obviously broken, bail ... */ > + if ((count - 4) == i) { > + dev_err(priv->dev, "No sync sequence found, invalid bitstream."); > + err = -EINVAL; > + goto out_free; > } > > /* enable clock */ >