Re: [PATCH v3] usb: gadget: f_mass_storage: Make CD-ROM emulation work with Mac OS-X

[Jack Pham <quic_jackp@quicinc.com>]

On Thu, Jan 13, 2022 at 09:59:16AM -0800, Jack Pham wrote:
> From: Roger Quadros <roger.quadros@nokia.com>
> 
> Mac OS-X expects CD-ROM TOC in raw format (i.e. format:2). It also
> sends the READ_TOC CDB in old style SFF8020i format. i.e. 2 format bits
> are encoded in MSBs of CDB byte 9.
> 
> This patch will enable CD-ROM emulation to work with Mac OS-X. Tested on
> Mac OS X v10.6.3.
> 
> Signed-off-by: Roger Quadros <roger.quadros@nokia.com>
> Signed-off-by: Jack Pham <quic_jackp@quicinc.com>
> ---
> v3: Updated command mask to allow for non-zero byte 2
> v2: Removed Change-Id
> v1: Resurrected original change [1] and consolidated into single patch
> 
> [1] https://lore.kernel.org/lkml/1302015569-9668-1-git-send-email-roger.quadros@nokia.com/T/#u
> 
>  drivers/usb/gadget/function/f_mass_storage.c | 73 ++++++++++++++++----
>  1 file changed, 61 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/usb/gadget/function/f_mass_storage.c b/drivers/usb/gadget/function/f_mass_storage.c
> index 752439690fda..d08463756eb0 100644
> --- a/drivers/usb/gadget/function/f_mass_storage.c
> +++ b/drivers/usb/gadget/function/f_mass_storage.c
> @@ -1188,6 +1188,8 @@ static int do_read_toc(struct fsg_common *common, struct fsg_buffhd *bh)
>  	int		msf = common->cmnd[1] & 0x02;
>  	int		start_track = common->cmnd[6];
>  	u8		*buf = (u8 *)bh->buf;
> +	u8		format;
> +	int		i, len;
>  
>  	if ((common->cmnd[1] & ~0x02) != 0 ||	/* Mask away MSF */
>  			start_track > 1) {
> @@ -1195,18 +1197,65 @@ static int do_read_toc(struct fsg_common *common, struct fsg_buffhd *bh)
>  		return -EINVAL;
>  	}
>  
> -	memset(buf, 0, 20);
> -	buf[1] = (20-2);		/* TOC data length */
> -	buf[2] = 1;			/* First track number */
> -	buf[3] = 1;			/* Last track number */
> -	buf[5] = 0x16;			/* Data track, copying allowed */
> -	buf[6] = 0x01;			/* Only track is number 1 */
> -	store_cdrom_address(&buf[8], msf, 0);
> +	format = common->cmnd[2] & 0xf;
> +	/*
> +	 * Check if CDB is old style SFF-8020i
> +	 * i.e. format is in 2 MSBs of byte 9
> +	 * Mac OS-X host sends us this.
> +	 */
> +	if (format == 0)
> +		format = (common->cmnd[9] >> 6) & 0x3;
> +
> +	switch (format) {
> +	case 0:
> +		/* Formatted TOC */
> +		len = 4 + 2*8;		/* 4 byte header + 2 descriptors */
> +		memset(buf, 0, len);
> +		len -= 2;		/* TOC Length excludes length field */

Length is now decremented to 18...

> +		buf[1] = len;		/* TOC data length */
> +		buf[2] = 1;		/* First track number */
> +		buf[3] = 1;		/* Last track number */
> +		buf[5] = 0x16;		/* Data track, copying allowed */
> +		buf[6] = 0x01;		/* Only track is number 1 */
> +		store_cdrom_address(&buf[8], msf, 0);
> +
> +		buf[13] = 0x16;		/* Lead-out track is data */
> +		buf[14] = 0xAA;		/* Lead-out track number */
> +		store_cdrom_address(&buf[16], msf, curlun->num_sectors);
> +		return len;

... which is returned here.  This should be returning 20, not 18.
Indeed it's the same mistake as was pointed out in the original patch
series https://lore.kernel.org/lkml/4D876ECA.4010503@nokia.com/ so I
must have inadvertently applied the wrong version of the patch.

> +
> +	case 2:
> +		/* Raw TOC */
> +		len = 4 + 3*11;		/* 4 byte header + 3 descriptors */
> +		memset(buf, 0, len);	/* Header + A0, A1 & A2 descriptors */
> +		len -= 2;		/* TOC Length excludes length field */

Same here.

Will send out v4 with this corrected.