163 posts
Pennsylvania
1281 posts
local newPath = "/fakeroot/"
local overrideFs= {}
for k,v in pairs _G.fs do
overrideFs[k] = function(...)
local args = {...}
for i=1,#args do
if k == "open" and i ~= 2 then
args[i] = newPath..args[i]
end
end
return v(unpack(args))
end
end
setmetatable(getfenv(), {
__index = function(t,k)
if k == "fs" then
return overrideFs
end
return _G[k]
end
})
163 posts
Pennsylvania
1281 posts
163 posts
Pennsylvania
wrappedFS.lua:18: vm error:
java.lang.ArrayIndexOutOfBoundsException
1281 posts
setmetatable(loadfile"/fileName",
__index = function(t,k)
if k == "fs" then
return overrideFs[k]
elseif _G[k] and k ~= "_G" then
return _G[k]
end
end
})()
2151 posts
Auckland, New Zealand
163 posts
Pennsylvania
2151 posts
Auckland, New Zealand
I'll see if I can come up with a way to recursively run CometWolf's routine on every CC API. Then I'll never have to update unless they add a new hardcoded API.
7508 posts
Australia
why not just do it based on the list taken from fs.list(/rom/apis/) if you change the pointer to the FS API in all those to point to the new FS API then you'd be good.Yes, I'd be interested to see if you can, that will prevent the need to manually update OneOS every time APIs are changed or added.I'll see if I can come up with a way to recursively run CometWolf's routine on every CC API. Then I'll never have to update unless they add a new hardcoded API.
1281 posts
for k,v in pairs(_G) do
if type(v) == "table" then
setmetatable(v,
{
__index = function(t2,k2)
if k2 == "fs" then
return overrideFs[k2]
end
return _G[k2]
end
}
)
end
end
2151 posts
Auckland, New Zealand
Would it though? I found that the API's FS wouldn't change.why not just do it based on the list taken from fs.list(/rom/apis/) if you change the pointer to the FS API in all those to point to the new FS API then you'd be good.Yes, I'd be interested to see if you can, that will prevent the need to manually update OneOS every time APIs are changed or added.I'll see if I can come up with a way to recursively run CometWolf's routine on every CC API. Then I'll never have to update unless they add a new hardcoded API.
163 posts
Pennsylvania
Would it though? I found that the API's FS wouldn't change.why not just do it based on the list taken from fs.list(/rom/apis/) if you change the pointer to the FS API in all those to point to the new FS API then you'd be good.Yes, I'd be interested to see if you can, that will prevent the need to manually update OneOS every time APIs are changed or added.I'll see if I can come up with a way to recursively run CometWolf's routine on every CC API. Then I'll never have to update unless they add a new hardcoded API.
local newRoot = "/t/"
local apiDir = '/rom/apis/'
local tAPI = fs.list(apiDir)
local oldFS = fs
for n,e in ipairs(tAPI) do
if not oldFS.isDir(apiDir..e) then
print('Unloading '..e)
os.unloadAPI(apiDir..e)
end
end
function loadAPI(path)
local name = string.match(oldFS.getName(path), '(%a+)%.?.-')
local env = setmetatable({fs = overrideFs, os = overrideOs}, { __index = _G }) -- passes a ref of the shell api to the loaded api
local func, err = loadfile(path)
if not func then
return false, printError(err)
end
setfenv(func, env)
func()
local api = {}
for k,v in pairs(env) do
api[k] = v
end
_G[name] = api
return true
end
local overrideFs= {}
for k,v in ipairs(_G.fs) do
overrideFs[k] = function(...)
local args = {...}
for i=1,#args do
if not (k == "open" and i == 2) then
args[i] = newRoot..args[i]
end
end
return v(unpack(args))
end
end
local overrideOs= {}
for k,v in ipairs(_G.os) do
overrideOs[k] = function(...)
local args = {...}
for i=1,#args do
if k == "loadAPI" or k == "unloadAPI" then
args[i] = newRoot..args[i]
end
end
return v(unpack(args))
end
end
setmetatable(getfenv(0), {
__index = function(t,k)
if k ~= "_G" then
if k == "fs" then
return overrideFs[k]
elseif k == "os" then
return overrideOs[k]
elseif _G[k] then
return _G[k]
end
end
end
})
for n,e in ipairs(tAPI) do
if not oldFS.isDir(apiDir..e) then
print('Loading '..e)
loadAPI(apiDir..e)
end
end
shell.run('ls') -- in here to make sure shell.isDir() functions properly
1610 posts
163 posts
Pennsylvania
Good point. But I might leave that up to OS authors. They may want a root mode, and it's not really my place to restrict them. I may find a way to require a password, though.Don't forget, if they pass the path ".." They're back in the root
1610 posts
Tip: GitHub repo apemanzilla/ApeOs, then navigate to /system/apis/safe_FS.luaGood point. But I might leave that up to OS authors. They may want a root mode, and it's not really my place to restrict them. I may find a way to require a password, though.Don't forget, if they pass the path ".." They're back in the root
163 posts
Pennsylvania
I looked at how it's implemented, but it doesn't look like it overrides fs. Do you do that elsewhere?Tip: GitHub repo apemanzilla/ApeOs, then navigate to /system/apis/safe_FS.luaGood point. But I might leave that up to OS authors. They may want a root mode, and it's not really my place to restrict them. I may find a way to require a password, though.Don't forget, if they pass the path ".." They're back in the root
Should give you a pretty well sandboxed fs API. (It also include io.open :P/>)
7508 posts
Australia
Good point. But I might leave that up to OS authors. They may want a root mode, and it's not really my place to restrict them. I may find a way to require a password, though.Don't forget, if they pass the path ".." They're back in the root
local path = root..shell.resolve(filePath)
163 posts
Pennsylvania
I tested his hypothesis, and this worked:Good point. But I might leave that up to OS authors. They may want a root mode, and it's not really my place to restrict them. I may find a way to require a password, though.Don't forget, if they pass the path ".." They're back in the rootfilePath can contain as many .. as they want, or any forwards slashes or anything, but they're not getting anywhere with it.local path = root..shell.resolve(filePath)
the best thing about shell.resolve is that the file does not need to exist for the file path to be sanitised :)/>
> test2
> cd ..
..> ls
rom
test test2
..>
7508 posts
Australia
what?I tested his hypothesis, and this worked:> test2 > cd .. ..> ls rom test test2 ..>
163 posts
Pennsylvania
I was able to cd into '..', or root, in this case.what?I tested his hypothesis, and this worked:> test2 > cd .. ..> ls rom test test2 ..>
7508 posts
Australia
after having implemented the code that I suggested?I was able to cd into '..', or root, in this case.
There was nothing preventing me from doing so.
1610 posts
163 posts
Pennsylvania
Yeah, since I'm on Linux, '..' is actually a symlink to the parent directory. On windows however, it's a special case handled by cd, so I think your code would work there, but since it's an actual directory on my fs, the only way to sanitize it correctly would be to check for and replace absolute paths equal to '/..'after having implemented the code that I suggested?I was able to cd into '..', or root, in this case.
There was nothing preventing me from doing so.
7508 posts
Australia
I'm on OSX I know what '..' is and its handled the same way as it is in Linux ('cause both are Unix). I performed tests and it worked perfectly fine for me. could I possibly get a copy of the code you used to test?Yeah, since I'm on Linux, '..' is actually a symlink to the parent directory. On windows however, it's a special case handled by cd, so I think your code would work there, but since it's an actual directory on my fs, the only way to sanitize it correctly would be to check for and replace absolute paths equal to '/..'
163 posts
Pennsylvania
Unfortunately, I've been overhauling my code, and I don't think the version that gave me that result exists.I'm on OSX I know what '..' is and its handled the same way as it is in Linux ('cause both are Unix). I performed tests and it worked perfectly fine for me. could I possibly get a copy of the code you used to test?Yeah, since I'm on Linux, '..' is actually a symlink to the parent directory. On windows however, it's a special case handled by cd, so I think your code would work there, but since it's an actual directory on my fs, the only way to sanitize it correctly would be to check for and replace absolute paths equal to '/..'
1610 posts
*Facepalm*Unfortunately, I've been overhauling my code, and I don't think the version that gave me that result exists.I'm on OSX I know what '..' is and its handled the same way as it is in Linux ('cause both are Unix). I performed tests and it worked perfectly fine for me. could I possibly get a copy of the code you used to test?Yeah, since I'm on Linux, '..' is actually a symlink to the parent directory. On windows however, it's a special case handled by cd, so I think your code would work there, but since it's an actual directory on my fs, the only way to sanitize it correctly would be to check for and replace absolute paths equal to '/..'