Fix <img /> detection regex in web cmdlets (#12099)

src/Microsoft.PowerShell.Commands.Utility/commands/utility/WebCmdlet/Common/BasicHtmlWebResponseObject.Common.cs

@@ -232,7 +232,7 @@ namespace Microsoft.PowerShell.Commands
 
             if (s_imageRegex == null)
             {
-                s_imageRegex = new Regex(@"<img\s+[^\s>]*>",
+                s_imageRegex = new Regex(@"<img\s[^>]*?>",
                     RegexOptions.Singleline | RegexOptions.IgnoreCase | RegexOptions.Compiled);
             }
         }

test/powershell/Modules/Microsoft.PowerShell.Utility/WebCmdlets.Tests.ps1

@@ -1900,7 +1900,51 @@ Describe "Invoke-WebRequest tests" -Tags "Feature", "RequireAdminOnWindows" {
         }
     }
 
+    Context "Regex Parsing" {
+
+        It 'correctly parses an image with id, class, and src attributes' {
+            $dosUri = Get-WebListenerUrl -Test 'Dos' -query @{
+                dosType = 'img-attribute'
+            }
+
+            $response = Invoke-WebRequest -Uri $dosUri
+            $response.Images | Should -Not -BeNullOrEmpty
+        }
+    }
+
     Context "Denial of service" -Tag 'DOS' {
+        It "Image Parsing" {
+            Set-ItResult -Pending -Because "The pathological regex runs fast due to https://github.com/dotnet/runtime/issues/33399.  Fixed in .NET 5 preview.2"
+            $dosUri = Get-WebListenerUrl -Test 'Dos' -query @{
+                dosType='img'
+                dosLength='5000'
+            }
+            $script:content = ''
+            [TimeSpan] $timeSpan = Measure-Command {
+                $response = Invoke-WebRequest -Uri $dosUri
+                $script:content = $response.content
+                $response.Images | out-null
+            }
+
+            $script:content | Should -Not -BeNullOrEmpty
+
+            # pathological regex
+            $regex = [RegEx]::new('<img\s+[^>]*>')
+
+            [TimeSpan] $pathologicalTimeSpan = Measure-Command {
+                $regex.Match($content)
+            }
+
+            $pathologicalRatio = $pathologicalTimeSpan.TotalMilliseconds/$timeSpan.TotalMilliseconds
+            Write-Verbose "Pathological ratio: $pathologicalRatio" -Verbose
+
+            # dosLength 4,000 on my 3.5 GHz 6-Core Intel Xeon E5 macpro produced a ratio of 12
+            # dosLength 5,000 on my 3.5 GHz 6-Core Intel Xeon E5 macpro produced a ratio of 21
+            # dosLength 10,000 on my 3.5 GHz 6-Core Intel Xeon E5 macpro produced a ratio of 75
+            # in some cases we will be running in a Docker container with modest resources
+            $pathologicalRatio | Should -BeGreaterThan 5
+        }
+
         It "Charset Parsing" {
             $dosUri = Get-WebListenerUrl -Test 'Dos' -query @{
                 dosType='charset'

test/tools/WebListener/Controllers/DosController.cs

@@ -29,33 +29,38 @@ namespace mvc.Controllers
             }
 
             StringValues dosLengths;
-            Int32 dosLength =1;
+            Int32 dosLength = 1;
             if (Request.Query.TryGetValue("dosLength", out dosLengths))
             {
                 Int32.TryParse(dosLengths.FirstOrDefault(), out dosLength);
             }
 
             string body = string.Empty;
-            switch(dosType)
+            switch (dosType)
             {
                 case "img":
                     contentType = "text/html; charset=utf8";
                     body = "<img" + (new string(' ', dosLength));
                     break;
+                // This is not really a DOS test, but this is the best place for it at present.
+                case "img-attribute":
+                    contentType = "text/html; charset=utf8";
+                    body = "<img src=\"https://fakesite.org/image.png\" id=\"mainImage\" class=\"lightbox\">";
+                    break;
                 case "charset":
                     contentType = "text/html; charset=melon";
                     body = "<meta " + (new string('.', dosLength));
                     break;
                 default:
-                    throw new InvalidOperationException("Invalid dosType: "+dosType);
+                    throw new InvalidOperationException("Invalid dosType: " + dosType);
             }
 
             // Content-Type must be applied right before it is sent to the client or MVC will overwrite.
             Response.OnStarting(state =>
                 {
-                     var httpContext = (HttpContext) state;
-                     httpContext.Response.ContentType = contentType;
-                     return Task.FromResult(0);
+                    var httpContext = (HttpContext)state;
+                    httpContext.Response.ContentType = contentType;
+                    return Task.FromResult(0);
                 }, HttpContext);
 
             Response.ContentLength = Encoding.UTF8.GetBytes(body).Length;