github.com/nicocha30/gvisor-ligolo@v0.0.0-20230726075806-989fa2c0a413/pkg/sentry/kernel/fd_table_unsafe.go (about)

     1  // Copyright 2018 The gVisor Authors.
     2  //
     3  // Licensed under the Apache License, Version 2.0 (the "License");
     4  // you may not use this file except in compliance with the License.
     5  // You may obtain a copy of the License at
     6  //
     7  //     http://www.apache.org/licenses/LICENSE-2.0
     8  //
     9  // Unless required by applicable law or agreed to in writing, software
    10  // distributed under the License is distributed on an "AS IS" BASIS,
    11  // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    12  // See the License for the specific language governing permissions and
    13  // limitations under the License.
    14  
    15  package kernel
    16  
    17  import (
    18  	"math"
    19  
    20  	"github.com/nicocha30/gvisor-ligolo/pkg/bitmap"
    21  	"github.com/nicocha30/gvisor-ligolo/pkg/sentry/vfs"
    22  )
    23  
    24  type descriptorBucket [fdsPerBucket]descriptorAtomicPtr
    25  type descriptorBucketSlice []descriptorBucketAtomicPtr
    26  
    27  // descriptorTable is a two level table. The first level is a slice of
    28  // *descriptorBucket where each bucket is a slice of *descriptor.
    29  //
    30  // All objects are updated atomically.
    31  type descriptorTable struct {
    32  	// Changes to the slice itself requiring holding FDTable.mu.
    33  	slice descriptorBucketSliceAtomicPtr `state:".(map[int32]*descriptor)"`
    34  }
    35  
    36  // initNoLeakCheck initializes the table without enabling leak checking.
    37  //
    38  // This is used when loading an FDTable after S/R, during which the ref count
    39  // object itself will enable leak checking if necessary.
    40  func (f *FDTable) initNoLeakCheck() {
    41  	var slice descriptorBucketSlice // Empty slice.
    42  	f.slice.Store(&slice)
    43  }
    44  
    45  // init initializes the table with leak checking.
    46  func (f *FDTable) init() {
    47  	f.initNoLeakCheck()
    48  	f.InitRefs()
    49  	f.fdBitmap = bitmap.New(uint32(math.MaxUint16))
    50  }
    51  
    52  const (
    53  	// fdsPerBucketShift is chosen in such a way that the size of bucket is
    54  	// equal to one page.
    55  	fdsPerBucketShift = 9
    56  	fdsPerBucket      = 1 << fdsPerBucketShift
    57  	fdsPerBucketMask  = fdsPerBucket - 1
    58  )
    59  
    60  // get gets a file entry.
    61  //
    62  // The boolean indicates whether this was in range.
    63  //
    64  //go:nosplit
    65  func (f *FDTable) get(fd int32) (*vfs.FileDescription, FDFlags, bool) {
    66  	slice := *f.slice.Load()
    67  	bucketN := fd >> fdsPerBucketShift
    68  	if bucketN >= int32(len(slice)) {
    69  		return nil, FDFlags{}, false
    70  	}
    71  	bucket := slice[bucketN].Load()
    72  	if bucket == nil {
    73  		return nil, FDFlags{}, false
    74  	}
    75  	d := bucket[fd&fdsPerBucketMask].Load()
    76  	if d == nil {
    77  		return nil, FDFlags{}, true
    78  	}
    79  	return d.file, d.flags, true
    80  }
    81  
    82  // CurrentMaxFDs returns the number of file descriptors that may be stored in f
    83  // without reallocation.
    84  func (f *FDTable) CurrentMaxFDs() int {
    85  	slice := *f.slice.Load()
    86  	return len(slice) * fdsPerBucket
    87  }
    88  
    89  // set sets the file description referred to by fd to file. If
    90  // file is non-nil, it takes a reference on them. If setAll replaces
    91  // an existing file description, it returns it with the FDTable's reference
    92  // transferred to the caller, which must call f.drop on the returned
    93  // file after unlocking f.mu.
    94  //
    95  // Precondition: mu must be held.
    96  func (f *FDTable) set(fd int32, file *vfs.FileDescription, flags FDFlags) *vfs.FileDescription {
    97  	slicePtr := f.slice.Load()
    98  
    99  	bucketN := fd >> fdsPerBucketShift
   100  	// Grow the table as required.
   101  	if length := len(*slicePtr); int(bucketN) >= length {
   102  		newLen := int(bucketN) + 1
   103  		if newLen < 2*length {
   104  			// Ensure the table at least doubles in size without going over the limit.
   105  			newLen = 2 * length
   106  			if newLen > int(MaxFdLimit) {
   107  				newLen = int(MaxFdLimit)
   108  			}
   109  		}
   110  		newSlice := append(*slicePtr, make([]descriptorBucketAtomicPtr, newLen-length)...)
   111  		slicePtr = &newSlice
   112  		f.slice.Store(slicePtr)
   113  	}
   114  
   115  	slice := *slicePtr
   116  
   117  	bucket := slice[bucketN].Load()
   118  	if bucket == nil {
   119  		bucket = &descriptorBucket{}
   120  		slice[bucketN].Store(bucket)
   121  	}
   122  
   123  	var desc *descriptor
   124  	if file != nil {
   125  		desc = &descriptor{
   126  			file:  file,
   127  			flags: flags,
   128  		}
   129  	}
   130  
   131  	// Update the single element.
   132  	orig := bucket[fd%fdsPerBucket].Swap(desc)
   133  
   134  	// Acquire a table reference.
   135  	if desc != nil && desc.file != nil {
   136  		if orig == nil || desc.file != orig.file {
   137  			desc.file.IncRef()
   138  		}
   139  	}
   140  
   141  	if orig != nil && orig.file != nil {
   142  		if desc == nil || desc.file != orig.file {
   143  			return orig.file
   144  		}
   145  	}
   146  	return nil
   147  }