Re: [9front] [PATCH] Unmount to remove sharp devices.

[Jacob Moody <moody@mail.posixcafe.org>]

On 5/26/22 17:36, unobe@cpan.org wrote:
> Great work on this moody.  I have some (hopefully helpful) questions
> below.
> 
> 
> Quoth Jacob Moody <moody@mail.posixcafe.org>:
>> Another iteration.
>> To state all of what this does:
>>
>> A process can remove access to a driver through writing
>>
>> chdev [ 0! ] devmask
>>
>> to /dev/drivers. This is tied to the processes namegroup,
>> and is inherited on both RFCNAMEG and RFNAMEG. In order to provide
>> all the knobs RFNOMNT does, blocking 'M' is interpreted to block all mounting.
>>
>> Support for this new operation has been added to namespace files,
>> along with a matching chdev(1). /lib/namespace.ftp has been updated
>> to use chdev.
>> ...
>> --- /tmp/diff100000711164
>> +++ b/rc/bin/chdev
>> @@ -1,0 +1,3 @@
>> +#!/bin/rc
>> +
>> +echo chdev $* >> '#c/drivers'
>> ...
>> --- /tmp/diff100000711170
>> +++ b/sys/man/1/chdev
>> @@ -1,0 +1,93 @@
>> +.TH CHDEV 1
>> +.SH NAME
>> +chdev \- change kernel driver access
>> +.SH SYNOPSIS
>> +.B chdev
>> +[ 0! ] \f2devmask\fP...
> 
>> --- a/sys/src/9/port/dev.c
>> +++ b/sys/src/9/port/dev.c
>> @@ -31,6 +31,74 @@
>>  }
>>
>>  void
>> +devmask(Pgrp *pgrp, char *devs)
>> +{
>> +	int i, t, w;
>> +	int invert;
>> +	char *p;
>> +	Rune r;
>> +	u64int mask[nelem(pgrp->notallowed)];
>> +
>> +	
>> +	invert = 1;
>> +	switch(*devs){
>> +	case '!':
>> +		memset(mask, 0, sizeof mask);
>> +		devs++;
>> +		invert--;
>> +		break;
>> +	case '0':
>> +		devs = "";
>> +	default:
>> +		memset(mask, 0xFF, sizeof mask);
>> +		break;
>> +	}		
>> +
>> +	w = sizeof mask[0] * 8;
>> +	for(p = devs; *p != '\0';){
>> +		p += chartorune(&r, p);
>> +		t = devno(r, 1);
>> +		if(t == -1)
>> +			continue;
>> +		if(invert)
>> +			mask[t/w] &= ~(1<<t%w);
>> +		else
>> +			mask[t/w] |= 1<<t%w;
>> +	}
> 
> What happens if '!!!' is the prefix?  Should the '!' set invert = 0
> explicitly instead of invert--?  Does '!0' allow access to all
> drivers?
> 

That switch case is not looped over. We only check the very first character
for the prefix. The rest of the string is interpreted to be the driver string.
In both of your cases the prefix is '!' and the driver string only contains '0' or '!',
neither are real drivers so you block nothing.